When you typed who is on snl tonight? You used to get 10 pages of clickbait that made you scroll down endlessly for the answer. The best product for most questions is to just give you the answer. That said, the death of discovery is worth mourning. I hope someone solves it but this was dying before ai came along. SEO killed the web, AI killed SEO, and so it goes…
I want to argue against you, but I just can't. Beyond the hyperbole, maybe we are on the cusp of change. Stuff being posted is AI slop, but maybe it is the Geo Cities obnoxious site of 1999, when an person does it because they want it and not a content farm.
Obviously, I'm not thinking about "efficiency" and "productivity gains" here, but I do have fond memories of ordering Chinese food over the phone. We had a few places and their phone numbers on the fridge, like that takeout place, or places like Pizza Hut (the dine-in experience was awesome!), and so on.
It felt like places had soul, even the Taco Bell or Pizza hut; now everything is corpo-slop or ghost kitchens.
I'm far happier making my own food at home, and I have zero interest in "supporting" local establishments that want to shake me down because they are run like PE sweat shops.
> It felt like places had soul, even the Taco Bell or Pizza hut; now everything is corpo-slop or ghost kitchens.
Yeah, I don’t understand it but so many casual restaurants do not want you eating in them anymore, and it’s not just the big chains.
It’s confusing to me. Does nobody else have business or
a fun activity in another neighborhood and stop for a quick bite anymore? Or meet up with friends at the cheap restaurant down the street?
Might it be that the restaurants are responding to society not wanting to eat in them?
I recently moved back to a city I lived in before Covid, and the biggest shock is how most eateries survived the pandemic, but they have few actual patrons now, only a constant flow of square-backpack delivery drivers going in and out of them. I'm told it's because people in their 20s and early 30s now (i.e. the demographic here with a lot of disposable income) aren't much interested in going out to eat; it has become normal that the food comes to them.
The difference is moderation and the sheer amount of bots swarming everything. You can't put anything up a website with user input without paying for 15 different services to shield you from the hive. A website for people to talk to each other? You'll be arrested after it turns out adults use it to talk to minors. You want people to post images? Oh god I hope you're paying for state of the art content moderation LLMs, or risk traumatizing your most dedicated users who volunteer.
There's opportunity out there to build the tools that enable protection for smaller, more focused internet. Until then, its a dark forest.
I don’t intend to argue against your main point but for anyone who it might be helpful to, OpenAI has a free moderation API. A few years ago moderation was either something you did by hand or got volunteers to do.
> A website for people to talk to each other? You'll be arrested after it turns out adults use it to talk to minors
Adults talking to minors of the forum wont make you arrested. But, blatantly facilitating child abuse again and again and again maybe with some low probability will.
Eh, there are plenty of forums out there doing just fine, and the same with other sites like wikis, image boards, chat rooms, etc.
None of them have needed to sign up for any sort of AI image classification service or moderation setup. My own site has done fine for the last decade with just human moderators and a relaxed community.
The personal web. Public entities, like governments, companies and non-profits still put in the effort to run their own websites, of which there are millions.
It's hardly surprising most normal private individuals don't want to do this and prefer a Web 2.0 style hosted app. They offer better convenience, design, privacy, security, etc.
I feel it's pronounced online. I'm a big gossip when it comes to new businesses coming into the neighborhood. I notice a tenant goes out and then I'm watching and asking people if they know anything about what's coming in. Everyone I know seems to be doing something similar. We may not visit if we're not interested in the goods/services, but we're aware of it if the need arises. But websites launch in the dark, practically invisible. If I search for it's exact good/service, no chance I'll find it (even pre-AI). It's pretty much a field of dreams with a sad ending, or a whole lot of work and money to buy traffic. Even some of these big Show HN's that spend a day or so on the front page, I bet most of them go back to ~0 real fast after that. There was certainly a time where that was like a winning lottery ticket. Maybe I'm off the mark, or cynical, I'd love to hear a story of a recent example where Show HN turned into a big MRR in a short timespan.
I forget the name but there once was a short lived search engine that heavily leaned on crawling new domains.
I only ran a dozen queries but it was memorable. I didn't see a website with "under construction" in it in the decade before nor the decade after.
It seems like an unfinished approach. (Perhaps add dns data?)
Someone should do a p2p search engine and crawl those 100 to 800 thousand new domains per day. Leave them in the index for a week or a month. Make the removed ones into a download, perhaps with some [traffic] data.
Then have the user set up some queries and give them a notification if a new result happens. With suggestions like `$city $restaurant`
Problem is, nobody cares enough to do that. To use that, I mean. These things were neat in the earlier Internet when it was a toy and we were seeking information. Now we have too much information.
The whole StumbleUpon idea was huge for a while too. Didn’t age well though, it was a novelty. I think this was around the time most people switched to a Feed style of information and that sucks your available attention
The once beautiful visualisation of nodes has turned into a glowing donut of activity. Lots of people care enough to use it. You don't have to use the network, it's already great if you only crawl the website you care about.
Ideal I think is to give it some bandwidth the first few days then tune it down to the minimum.
Actually, it isn't. We should have many protocols in the browser.
With enough gate keeping things could get to a point where the peasants have to extend a document format to be able to publish applications.
When that happens we will have to prevent users from finding them until we find a better method of shutting them down.
One day we will control what they can talk about and our official narrative will be all that remains.</sarcasm>
My sector is currently on strike, I could find exactly one outdated news article where one of the big employers says it has no influence on operations. Like they would pay many millions per month for work they could easily do without.
Is writing happening anywhere? That's my big fear.
The last two surviving text based sites are reddit and x. Reddit is going out of their way to become an Instagram feed, and what remains is mostly bots and astroturfing. X, even ignoring the bots and nazis, is not a place for long format.
There's still pockets like IMDb reviews or this site, but it's an extreme niche. I even have people at work who communicate by audios and videos whenever they can rather than writing on slack.
There will eventually be small, highly moderated corners of what was once an open internet pop up. They probably already exist. Nothing of this internet serves people, the people serve it. That will change eventually.
I was about to mention this and noticed your post. Not my cup of tea, but my niece was holding her breath for it to come back up for the full two and a bit weeks it was down. I told her to prepare for the worst, that length downtime is usually fatal, but I think it was the day after I said that when it came back online and a great weight was lifted from my niece's shoulders. She's Gen Z and loves the written word. Little darlin'.
Mostly in its own domains. My niche interests are covered in sites their readers pay for, so i built a search engine to find such. Quite many sites supported with Patreon, Substack, Ko-fi etc. Shameless plug: readerfunded.com
The index is not very big, but it helps me find interesting content.
Google is responsible for adsense and they decide who gets search traffic and who doesn't. They are complicit in the enshittification of their own game.
They've made the internet broadly unuseable. I absolutely loathe visiting unfamiliar websites. Sometimes you can't even find the article amidst the spam, pop-ups and autoplay videos. Remember when we used to shame users for commenting before reading the article? Most of these sites today you wouldn't wish upon your childhood bully, of course nobody reads the articles anymore.
Meanwhile I've conditioned myself to not even recognize what it is that's being advertised at me, so it's all for naught.
It's a futile arms race between advertisers finding ever more annoying ways of getting my attention and me finding ever more devious ways of averting my eyes from whatever it is they want me to recognize.
I suspect Google drives just enough traffic to crappified websites like Pinterest, Yelp and IMDB to keep them on life support so it can avoid anti-trust lawsuits. That's the only reason they exist.
well.. that really wasn't my experience. when you put "who is on snl tonight," you would get a top row of news stories related to that, probably in this case some (admittedly low effort) post saying "Catch Tom Holland on SNL this Saturday!"
that said, it's true that SEO did kill even that, because the headlines began to be "You won't believe who tonight's SNL headliner is" and they even keep the name out of the slug too.
# You won't believe who tonight's SNL headliner is.
## Someone is hosting SNL this week.
In this article we will tell you who is hosting SNL this week.
### What is SNL?
SNL, or Satruday Night Live, is a sketch comedy show that originated back in...
### A brief history of sketch comedy
...
### Past SNL hosts
...
### This week on SNL
<this section still does not say who is hosting this week>
...
...
...
...
---
This almost every webpage/blog post I've found through a search engine since around 2016 (I remember around that time every single time I'd search something about React, every article would start with many paragraphs describing what React is).
Anyway, ya. People say that nostalgia for the "old web" is just that (nostalgia). I disagree, but ya.
Is there anything after AI though? As long as we allow AI to benefit individual people, we seem headed to an ad-free custom GUI future.
As you say, it has already started with AI rewriting web results into a focused text response--a text interface.
Frontier AI models have started drawing little diagrams and creating little interactive apps that go along with whatever they're trying to demonstrate. They also extract images from--well, from somewhere, the web I assume?--and placing those images in their text responses.
I've been using AI to write custom widgets for the game Beyond All Reason (which is open source and allows custom widgets), and it's great even with this obscure API. It can turn ideas into beautiful custom graphics easily.
I don't want to see your ads, or even your webpage. I just want the information.
Maybe with AI everywhere we can finally do some sort of micropayments where the AI can distribute a fixed amount of money on my behalf. I am willing to directly pay a small amount for casual browsing.
IMO it's a downstream problem of computers shifting from "specialized tool that responds to the owner and gives them agency" and towards "something effectively owned by a bunch of competing companies to push whatever they decide into your face."
The latter can be "thinned" much more effectively.
Then I guess this whole "free market" experiment is pretty much over, because aren't there several companies who offer flights? And those companies still want to attract customers right (I mean, if companies don't want customers, what are we even doing?), so why wouldn't other airlines offer information about their flights my AI can collect?
There may be a niche case where only one airline offers an unusual flight, so be it, but in general there are always going to be multiple companies offering flights, and each of those companies wants to participate in the market (that's kind of the point of our economic system) and so they will make information about their flights available.
You're right, but again, if people want to find flight information from multiple companies, but no company is willing or able to offer customers what they want, then why do we even have companies?
If push comes to shove given local AI it might well turn out that we don't need search engines to the same extent that we used to. Or rather that much of the process can be run locally.
Of course that would burn through a lot of resources and generally be inconvenient in various ways so inevitably someone would launch a centralized version and the cycle repeats.
> so why wouldn't other airlines offer information about their flights my AI can collect?
Because they'll want you to use their app to get that information so that they can collect massive amounts of your personal data and customize what flights and prices they'll offer you depending on who they think you are and how much money they think you have.
The "free market" means that companies are free make you use proprietary AI that shows you ads to access information about their goods and services. It also means that companies are free to make deals to collude about this. Same as how ticketmaster has deals with all the venues that says they have to use them for ticket sales. They are free to screw you over.
The only way to strip all ads is to deliver no information. They're not going to all be as obvious as "If you're hungry, why not try the new Bacon BaconBacon from Wendy's?"
Imagine you ask "quick snacks for Superbowl Sunday" - but the dairy industry has bought ad space. You'll get suggestions for cheesy nachos and spinach dip for sure, and chicken wings and other non-dairy options might be slightly less likely to appear, or described in slightly less enthusiastic tones.
And if the AI folk have their way, it won't stay in the chatbot either. When your day-planning agent or personal shopper gets ads, you won't even see the text.
Sorry for offtopic, at his point I more surprised (or just didn't notice) why there no local llm projects with browser plugin to filter/categorize HN content with headless entering site to analysis content. for example : topic with links, papers and algorithms inside, DIY style (no vendor locking) tutorial and so on.
1. "Ads" through LLMs are so much easier to personalize to the viewer and can be deployed at just the right moment to influence them in a way that might pass as organic.
2. Companies/politicians/wealthy people can simply pay for a model to be generically favorable to them, where positive words/events are mentioned and negative aspects are given dynamically-generated excuses or buried. (Or unfavorable to opponents, if the model-company thinks they can avoid lawsuits over it.)
3. Broad bias is easier to purchase, where a consortium or political party can pay to favor one entire class of thing over another, without developing individual ads for specific products or candidates.
In the end, the risk-profile is similar to hiring a human "personal assistant"... who takes secret bribes to manipulate you.
1. Do you have the right hardware? Most people don't and hardware has just exploded in price. One of the RAM makers and also AMD have said they will focus on data centers, which will only make things worse.
2. Do you have the know-how? It needs to be as easy as installing and using Chrome.
3. The end result needs to be directly as usable as ChatGPT.
I'm half optimistic about this, but we're so far from it from a hardware point of view.
We need unified memory for almost everything for these types of setups to be usable and we also need to standardize LLMs, if everyone wants to run their own LLM locally, that will probably be untenable for decades. We probably need new DRAM tech (3D RAM?).
Hardware prices need to go down a lot, and we're probably 5+ years away from that. Not only do prices need to go back to where they were, they need to become even lower. So we're probably 10+ years away from usable good enough local LLMs for the masses.
It's going to be an ugly decade from a hardware economics point of view. The potential is endless, but the way there is hard.
My dude, I get where you're coming from, but we've got decades of disappointing evidence that things will go a different way, absent something new like a political movement with legislation.
The predominantly average user isn't going to even know what the difference is between local and remote models, they're going to use whatever promises "fast and instant" access to read and modify the data that was already captured by "cloud" providers.
And that's not even starting to get into issues like figuring out which local model is actually trustworthy, picking something suitable to hardware available, how to hook up all the read/write access, etc.
I'm sure some people will try to game LLMs to get it to promote what they want, but it's only a matter of time until the AI companies offer manipulation and ad injection as a paid service.
> When you typed who is on snl tonight? You used to get 10 pages of clickbait that made you scroll down endlessly for the answer.
Right. Why? That's a nightmare that Google built for us. They lined the pockets of SEO spammers, and because it was their business, they talked the talk but were never willing to truly deal with the problem.
But even in your SNL example, to show you the answer, Google AI needs to take it from someone who actually made an effort to put this information online. Network programming is an rare corner case where the person doesn't mind the bot slurping it up because the goal is to get people to watch SNL, not to visit a website. But there is so much non-commercial content on the internet that is really getting hosed right now, and that people probably simply won't be putting online anymore because they don't want to do free labor for OpenAI or Google.
As a practical example, there's a lot of hobby historian communities dedicated to specific places, periods, or types of collectibles, and the only reason they exist is to bring people together and share their finds and wisdom. If you're a historian of Kinder Surprise toys or old military uniforms or whatever, and Google is now taking your work without letting you connect with like-minded individuals, there's really little incentive left.
Well, I'd say if people want communities, they'll always find a way to build connections—that part hasn't changed anyway.
The real difference comes down to who's paying the bill for all the server stuff. "Clicks" bring in ad revenue, which is what kept the internet "free" all these years, but that model might be coming to an end.
Those niche special interests will always have hangout spots. YouTubers putting out weekly videos, subreddits, discord servers, facebook groups, irc channels, conventions. As long as it’s something popular enough to have a certain critical mass —- maybe one thousand dedicated fans world wide —- there’s going to be places to hang out.
Who cares if it isn’t googleable. There’s other ways to build community and other avenues to be recognized for your work.
No ? They never had hangout spots, or found critical mass, until the internet came along and made finding people wildly easier.
AI search results are inserting themselves as intermediaries in that process of discovery. This is the opposite of what the net achieved, and its reducing the rate at which potential group members ever find community.
YouTube in particular is already getting hemorrhaged by ai content.
I don't think this will play out as you imagine there, because those platforms will drown out human made content with ai copycats until nothing but the already established players are left
Its not that hard to avoid. Whenever you see an AI screenshot, click dont recommend this channel again. After a few days they all clear out and you can go back to the traditional endless feed of nothing but low quality human made garbage that has nothing to do with your interests, just as God intended.
AIs are pretty good and helping you avoid AI content.
The search box on YouTube is worthless and searching from Google or Bing is almost as bad. But if you ask Claude for a non-AI video that explains X, it does a good job of finding it.
The real pro-tip though is to use YouTube with intention. Don't go there and start scrolling. Go there for something specific and leave as soon as the video is over.
It's funny. Until a couple years ago I really hated walled gardens like facebook groups, discord and other app-only communities. But now I find myself defending them, as a last frontier against automation, where you can still hangout with real humans online...
Depending on the group, because you know some personally who know others personally ... a chain of trust. In other cases because you at least know some online long enough to be quite sure they are human. Also most bots are still spottable, just not always on first glance.
The difference is that Google aggregates content from everywhere, so you don't need a pre-existing relationship with any of those places to find content. If you want to find a YouTube video without Google you have to go to YouTube to search for it. That's great for YouTube, but it makes it virtually impossible for any other video platform to get a foothold.
The beauty of Google is (was?) that it was a way to get traffic if you weren't already the incumbent. What you're suggesting is that we should give up on having new platforms and just stick with the existing ones, even if they're terrible now or they decline in the future. That's not fun.
Google wasn’t that important for the web. Plenty of the things I’ve learned about for the first time in the last decade haven’t been driven by Google, because the top Google results are mostly SEO nonsense.
Google was great for finding products and services, but poor at finding the type of communities we are describing.
The window where Google was an amazing way to find anything on the web was, at this point, quite small, on the other hand the things like AMP, killing off Google Reader, SEO and promoted listings were all detrimental for the web. So important, yes, but a net positive, I'm not sure anymore.
In my personal experience Google’s ability to find communities started declining about 2010 or so, when all of the web forums started migrating to Facebook groups and Subreddits.
Indeed. I average 1k page views/day, I would guess 300-500 visitors/day. I'm fine with that. For one thing, I can personally respond to every comment usually in real time.
Google directly and indirectly benefits from SEO spam. They are a solution to a messy web, the messier the web, the more necessary they are. Without web spam, Google would have just been another search engine next to Altavista.
Google is also in the unique position of selling both search ads and display ads. And they know what a visitors ad exposure is for links in the search results. They make money on both sides of the click.
SEO is not a general messy web though, it's a specific attack on Google and Google search users that makes Google search less useful. I think it's quite a stretch to say Google benefits from it.
> If you're a historian of Kinder Surprise toys or old military uniforms or whatever, and Google is now taking your work without letting you connect with like-minded individuals, there's really little incentive left.
Won't the people who are interested in connecting with like minded individuals click through anyway? The 40% who aren't clicking now are the people who were just coming to get the info and bounce, they weren't the ones connecting with like minded individuals.
If the sites go under because of this, there won’t be a site to click to in the first place.
Sites that were too big to just be run as a couple hours a week hobby project level, for years, relied on all traffic, including people who only read a page or two, to generate money via ad revenue. And the site being big, and existing in search results meant it was there for that core audience of people who might build community there.
All this is undone if hosting a site is no longer rewarding because you get no actual visitors and no community forms, while your content is just repackaged by commercial AI.
And I say this as a definite net promoter of AI. I just am honest that from a content publisher’s point of view, yup, they are getting screwed hard here - there’s no upside for them.
As an HN oldtimer, I can recognize your post as a typical reply made 10 years ago already, and as we can see, that never went anywhere. The web continued to decline.
The biggest problem is the general shift by which most of the world uses smartphones to access the internet, not computers. This medium seems to discourage traditional web content at a societal scale. A lot of younger people barely ever open a browser, their internet use is exclusively through apps. That causes traditional web content to gray out and become even less appealing to audiences.
>... smartphones to access the internet, not computers.
I was surprised when I learned that my obscure blog (no app; 1k page views/day average) gets 70% of its traffic from smartphones, congruent with your observation.
Since I solely use my laptop to produce my blog I was taken aback and immediately checked to see how it looks on my phone — something I'd never done.
I was relieved to see it displays perfectly there: in fact, it looks much better.
Most blogging platforms have had decent mobile theming for long years now. All the same, the small format of a smartphone screen seems to discourage, on a societal level, longform text reading, hence the expansion of microblogging, then image content in the Instagram age, then short-video content.
>... the small format of a smartphone screen seems to discourage, on a societal level, longform text reading...
FWIW just for lulz the other day I tried the iPhone Kindle app to see what it was like to read books there. I used white type on black background: excellent! Easier, sharper, all around better than my latest gen Kindle Paperwhite, which is what I use to read in bed at night.
I won't use screens* (read phones/tablets/laptops) in my bedroom so I'll stick to the Paperwhite there.
Yes, the Paperwhite has a screen, but e-ink is a whole different genre than the others.
This perhaps puts us in a situation where we return to the days of web 0.1... People put things online because they want them to be accessible online and for no other reason. Not for ad revenue, not for eyeball clout... Just because it's more convenient to have it online than stuck offline.
> People put things online because they want them to be accessible online
And the costs of putting something on line were commensurate with the number of people who needed to be paid to accomplish that goal. When Al Gore made the first website in the early 90s, it probably cost a few pennies of electricity and the effort of some grad students who worked for a few more pennies. Today, putting a website online involves paying a lot more companies who have massive bottom lines to support.
It really doesn't. No feature that Al Gore relied upon has been removed (with the possible exception of the blink tag).
I maintain a blog from a static served page. It's a fancy text file. Granted, it took me years of study to know what to put in the text file and I spend like a hundred bucks a year on hosting. But I spend more than that on a Netflix subscription.
You're thinking about putting a website online with a massive amount of functionality on top of it, or SEOing it, or monitoring it for traffic or ad integrating it or... All of that is web 1.0 and Web 2.0 thinking.
Web 0.1 was hand uploading bespoke HTML and possibly using a static site generator. Documents served by academics and government functionaries to other academics and government functionaries because the alternative ways to share those documents were more expensive.
The problem is most such sites were so horribly overloaded with obnoxious advertising that they were unusable without an ad blocker. They had already chosen to make their sites unbearable long before AI came along, so it's hard to mourn them now.
You are not born with interests, you develop them over time. Often by discovering that there exists a cool community you can be a part of.
And that's before we get to the other problem: what's under the AI summaries, and what's often "cited" by the summaries, is LLM slop too, because it's now practically free to create on any topic. For example, I searched for info on a vintage calculator just yesterday, and ended up on this page:
It's just data scraped from real hobby websites to boost their page rank. And everything is like that now. No subject is too insignificant for LLM spam because the cost is negligible.
I’m sorry but there is no one needed to post who is on SNL tonight, nor when certain sporting events are occurring, nor when the next episode of $SHOW comes on.
These are far more common than niche hobby collectible communities, and if those communities love bringing people together those people can still hang out together!
People who are looking for others with a similar passion—which, if it’s a niche interest, sounds quite nice—still have access to search results or the sources shown on the AI summary.
But for the vast majority of things, we don’t need 29 sites competing for SEO on “Which channel is the US Open on?”
I think there is probably some long-tail data, but I'm skeptical a large volume of questions actually ended up going to hobbyist historian communities in the first place. Unless I'm misremembering the before times, like 99% of my history-related queries on Google led me to Wikipedia. I might as well have been doing a site:wikipedia.org search for how reliably that happened. Now, the AI Overview just summarizes the same Wikipedia article. Not a huge change in that particular niche imo, unless you're tracking Wikipedia pageviews. Which even as a longtime Wikipedia editor I don't care about (I don't edit it to rack up pageviews).
Well that will at least undermine the point of Wikipedia, which is that readers can edit it and improve it. If people are reading it only through AI summaries, Wikipedia will eventually lose that critical mass of editor community that made it possible in the first place
> Well that will at least undermine the point of Wikipedia, which is that readers can edit it and improve it. If people are reading it only through AI summaries, Wikipedia will eventually lose that critical mass of editor community that made it possible in the first place
IMHO, Wikipedia has similar culture problems to Stack Overflow, just less extreme, and its editor decline trend started before AI and AI overviews.
Due to those culture problems, eventually Wikipedia will be completely dominated by people dedicated to pushing ideological agendas and others who just like throwing the book at newbies.
Most of those AI summaries are pulling right from Wikipedia and their answer accuracy is calibrated against it. If anything, the problem is going to be that Wikipedia answers become an unstopable circular source of truth.
2) reddit being plagued by bots, artificial upvotes
3) reddit mod problem (mods can create a community, but also mods cannot be removed. removal would be a problem due to brigading, but there is no mechanism to remove bad mods)
Wikipedia has never been a community of occasional contributors. It's controlled with an iron fist by a small group of insiders that treat the wiki pages like their personal fiefdom and often have no personal interest (or expertise) at all in the topics that they write about.
This may be the impression for people that are not familiar with how Wikipedia works (and that's a problem in itself), but in my experience (20+ years of contributions) this is not true at all. The community is very open to discuss different ideas and it always surprises me how opinions and consensus can quickly change with new information from reliable sources. As for the problem of seeming hostile to newcomers, this is due to the rules that govern the editing process in the community, the result of years of debates. This is in my opinion inevitable in order to run a successful community, but it also means that it requires some effort to join it
A good amount of Wikipedia pages are controlled by editors with autism (sometimes disabling to the point of inability to hold a job, hence so much time to edit Wikipedia), and indeed they have a personal interest in the subject that amounts to outright obsession--and that's the problem, that's why they can sometimes perceive other people's edits as a threat.
In this case it's the editors I interacted with (myself included) closely through long years of editing Wikipedia. Now, I moved on to other things before Covid, but with the difficulty of crowdsourced projects to ensure turnover of a healthy general public of contributors in the smartphone age, I can't imagine things have improved from this in any way.
Any most of them haven't done any real writing in years. They stand atop their domains angrily fighting off anyone who dares to update something without kissing the ring. They don't do research and only contribute through deletions and reversals of other people's writing.
I have a few relatives infamous enough to have Wikipedia entries. Every time I have tried to update them, a mod has pounced. Where my grandfather died ... nope. That cannot be changed without me submitting heaps of personal information to some rando mod. And good luck going anywhere near article about legal topics. Lawyer? Published expert in your narrow field? Not good enough. It's armchair lawyers only on wikipedia.
Wikipedia has problems with territorial editors, but you aren't going to get much sympathy when being rebuffed for editing pages based on personal acquaintance with the subject. That a higher standard of citation is required than "Trust me, I know, I'm a relative" has been a basic feature of the project since its founding in the early 2000s.
Which is a totally reasonable thing, but the amount of required evidence is arbitrary and dictated by angry mods. In this particular case, my grandfather had stepped out of public life decades earlier. So to prove that the "John Smith" who died in city X was the same "John Smith" who was famous in city Y, the mod wanted personal ID from me and an entire life history linking the famous "John Smith" to the dead one. Just... no. I am not going to draft a multi-page biography and hand over reams of family info to a rando username. My grandfather's one-page Wikipedia entry will just have to remain wrong, forever.
The other small change I wanted to make was adding a note stating that several of his listed children and grandchildren (ie me) had continued the family tradition of military service. Again, the mod wanted way too much PII. That small bit of color will also never appear in wikipedia. It is no wonder that famous/political families hire PR firms to negotiate and update their Wikipedia listings.
How is that any worse than citing some random newspaper? At some point _someone_ has to cite a primary source that isn't available online. Wikipedia is just kicking the can down the road.
An encyclopedia is a tertiary source. A random newspaper is a secondary source. It has always been the case that the former uses the latter as citations.
This reminds me of my experience some years ago with Reddit. I volunteered to host a real time Q&A about anesthesia (I'm a retired neurosurgical anesthesiologist). After about an hour of photographing and uploading every license/certificate of board qualification/evidence of continuing education requirements fulfilled/driver's license/utility bills/articles I've published/Google Scholar page/passport, they said my proof of identity/qualification was insufficient so my offer was denied.
It's very true. I am hesitant to post things I think are novel now because I know it will just go into training data, unattributed. I even comment on HN a lot less. It's wholesale ripping off of people's IP - the chilling effects of this will not be felt for a while.
Whether it goes into training data is irrelevant. The AI agent (whether embedded into google's search or otherwise) will do the search on behalf of the person, find the answer in your page and return the answer, rather than your page. It's the ability of AI to search and summarize that has changed things. Not whether the LLM itself has absorbed your particular information. Once LLM's were good enough to make reasonable tools calls to search the web it was game over. It's just like how calling a library to ask a factual question of a librarian, rather than having to locate the information in a primary (or secondary) source yourself disintermediated the encylopedia publishers. It's just that calling librarians was never particular efficient or practical.
Between SEO spam as the precursor then AI generated SEO spam and then just AI.
I used to enjoy stumbling on random things and learning/seeing into a world I didn't know existed (usually related to tech/science/history) but the signal/noise has changed radically and not in favour of the signal.
I think the problem is that we depend on Google or other mega-corporations
for information-discovery on the world wide web. It's not just a search
engine - it is now top-down control by these private entities.
We need a different version of the www. Google controls the www though via
the adChromium codebase.
The only actual good parts of the internet left are extremely walled off, gate kept places with a handful of people who know each other. Everything else is a completely lost cause, including this site.
Jeez, I didn't even think of providing free training data. I've been more worried about data harvesting and attributions to some sort of marketing profile (or worse).
>It's wholesale ripping off of people's IP - the chilling effects of this will not be felt for a while.
The wholesale ripping off started in the late 90's and we are already deep into the "feeling it". DRM, standalone apps for everything, consolidated corporate internet with paywalls all over.
I'm sorry, I just cannot take anyone on the internet who bemoans IP laws seriously. We all ad block everything and pirate stuff as needed, this is the moral high ground of the internet for 25+ years. Hell LLMs are the worst pirates anyway because they won't even let you just cleanly download the IP you want to steal.
Copyright is only useful for rent-seeking if you have a popular enough IP and lawyers on speed dial to collect that rent or for censoring the availability of IP. DNS blocks are a type of copyright censorship.
Copyright has little or no value for small blokes and bloketts like us. I think that part of the problem with copyright is that we are blending rent-seeking with acknowledgement of one's contribution in the creation of IP.
I dare say, most of the value of “copyright” for us small players is that it is part of the discovery process through which people with jobs and money find us. My writings and code have led customers to me (in a good way).
I don't mind what the LLM hovers up of my so-called creative efforts as long as people can find me. And that perspective is, I think, the core complaint people have about Google's AI searches. At least, that's how I read the article.
FWIW, having known the people responsible at Google for dealing with SEO spam...
It was never a question of resources and it was never a conflicting goals issue. From Google's standpoint, SEO spam is an existential threat because if somebody solves filtering for it, everyone will switch to that competitor.
It is simply a hard problem to solve at scale. Successful spammers live on the razor line between the signals Google has for useful content and wasting people's time.
The element Google can't control for is human nature. SEO spam sites are built to keep organic users reading long enough to tip the quality signal in the site's favor. Savvy users like those in this conversation know what they look like, but remember that most of the internet is not composed of sanity users. So the signal Google mostly gets from those pages is that they are holding user attention, and holding user attention, and holding use your attention because the site is annoying to retrieve information from is difficult to distinguish from holding user attention because the site is good.
I still keep my own website devoted to most of the interests I have, and sure, those opinions and information are destined for aggregation, but at least they are still worth something and reaching an audience, even if they are not the way I planned.
Once you've created something, whether it's an album or a blogpost, they are like your children - or as Kahlil Gibran wrote:
They come through you but not from you,
And though they are with you yet they belong not to you.
When people's interest is piqued, they may investigate more deeply, and that may lead them to other people in the same way. Discord servers are doing pretty well.
Personally, I'm a follower of Bowie's definition of success, in that success is about how much you influence other people, because it's purer and happens even if they hate you.
CREATING SOMETHING
__________________________________
.-' '-.
.' '.
/ \
/ \
| Curiosity |
| Joy |
| Art |
| Experimentation |
| Learning |
| Self-expression |
| Play |
| Discovery |
| |
| ___ |
| Making things .-' '-. |
| just because / MONEY \ |
| you want to \ / |
\ '-.___.-' /
\ /
'. _.'
'-.________________________________.-'
Monetisation is just one possible consequence of creation. We've somehow mistaken it for the purpose. Because capitalism.
Google could use the LLM to find the best hit in the SRPs and put those before the synopsis. It would take more context but if it has an idea of what the answer is it could review the text content of the top hits and check them against what it think the user is looking for.
Of course more often than not the reference links within LLM responses don't really relate to what they're saying - LLMs act like highschoolers, "I need X references to back up my essay, so I'll find some things that sound relevant and put them in as footnotes".
Almost nobody is using local AI. Almost nobody is using a computing device they actually control. Most people don't even have a desktop or laptop, only cell phones which further limits their options. Actual free local LLMs have the potential to be useful in a lot of ways, but it's only useful when they are powerful, with hardware requirements people can easily manage, on hardware/operating system they control. If you can't afford ram and your computer just logs you into a cloud account where your OS spies on everything you do and pushes ads your AI app isn't going to save you.
> You used to get 10 pages of clickbait that made you scroll down endlessly for the answer.
And before that, you would get the an exact, human written page for what you searched for, which is why Google was able to make money in the first place.
This is a Google search problem, and they could have defeated SEO sites if it weren't more profitable for them. SEOs didn't force Google to put 8 ads between the search bar and the first non-ad result, Google did.
Whatever Google is now, nobody would actually choose it if there were alternatives that competed on product, but there aren't.
We all live in duopoly land where every extremely high margin software and hardware role is taken up by 2 of 7 companies.
> Whatever Google is now, nobody would actually choose it if there were alternatives that competed on product, but there aren't.
There are dozens but they aren't good either.
It turns out that running a good search engine is very challenging and insanely expensive.
Companies make good products as loss leaders then monetize them later. People complain that the product was "enshittified" but what they get wrong is the original product never existed in any real sense. It was never a durable maintainable system. Most things require money to run well and if they can't make money they won't keep running well.
3. Threat of violence (usually via government / taxes)
You can have some mix of those three (a mix of 80% money 20% goodwill seems pretty solid in my experience), but I don't think there is some secret fourth option.
Google was born out of CIA and NSA research grants - US intelligence services saw the potential of the internet as a platform for mass surveillance, data collection and psychological warfare back in the early 1990s.
As with the CIA funding the modern art movement and various counterculture groups, and the LSD that drove hippie culture being a product of MKULTRA, the money would have appeared from somewhere.
We used to call that a bait and switch scam, yknow. You post as though it’s inevitable, Google was profitable before 8 ads per result. They didn’t have to cannibalize themselves, they chose to
The fact that google ads went from obvious and clear, to nearly indistinguishable from search results, and took up more screen space over time, has nothing to do with building a good product.
It has everything to do with quarterly number go up. The market is meant to align incentives, and the incentives here is to transfer as much benefit as is possible to shareholders.
Google (Tech in general) makes more money on a smaller cost base than any industry that has ever existed.
People say this and act like it has something to do with greed.
No. This is just basic economics. This is what it means to run a business with other people's money. If your growth doesn't compete with other places they could invest their money, why shouldn't they take it back?
> Google (Tech in general) makes more money on a smaller cost base than any industry that has ever existed.
This is quite simply not true. Tech does not have the largest profit margins of all time, and has exceptionally high capital requirements to operate.
What tech does have is really nice scaling properties, which Google benefits from as the "winner" in search. But beware survivorship bias.
I think tobacco is an industry that does better than tech.
Marginal costs of sales for tech are cheap, code scales like nothing that has ever existed before. I’ve worked in “legacy”, more capital intensive industries, where most of your revenue is consumed by your COGS. Tech has it absurdly absurdly good.
I am not sure if survivorship bias matters here; I can sort of see the value of the argument, but I am not entirely convinced.
It doesn’t change the operational logic of tech. Firms fail in all industries, so we are always looking at survivors when comparing. Competing is the basis of the market anyway, so the firms failing would be come under the cost of doing business.
I get reliable search results the vast majority of times, and if there’s a site that keeps popping up that I don’t like (AI-generated slop or paywalled), I block it and my results improve.
You can block sites, because Kagi’s product is search, not ads. Yeah, you have to pay for it, but you know that? I’m fine with that. I’m fine with paying for a product to do a thing well, and not have it tainted by the alternate revenue stream the company actually uses to make money from.
I've bounced off Kagi a couple times, but they've progressively improved as Google has worsened, and now I think I'm here to stay. They're dramatically better at skipping past literal whole pages of results of slop copypasta sites now. Google is rather clearly asleep at the wheel.
I think this kind of exemplifies the problem. Yes Kagi is a superior product. Why haven’t they taken off? It seems reasonable to conclude that competing on product isn’t enough.
Yeah google is deeply entrenched and people have come to expect things for free.
I use Kagi, and I really hope I am part of a large enough cohort paying for it to be a good, viable business. But realistically it is incredibly difficult to imagine it taking any major market share given the above.
Internet search isn’t one of those. Breadth, depth, and quality of search results (ie what you want is the first result) matter. That’s only reasonably attainable at scale. Not to mention it’d be a PITA to optimize for a dozen competing crawl standards.
Tell them, don't tell me! Yes, they are not doing much to reach low hanging fruit outside of tech industry people, even though a premium search engine has millions of potential customers.
I quit my Kagi subscription precisely because it's unable to keep up with the AI slop infesting every search. Even Kagi's results are mostly AI generated articles these days.
> And before that, you would get the an exact, human written page for what you searched for, which is why Google was able to make money in the first place.
Google wasn’t really good for questions like “who is hosting SNL?” until pretty late in the game. The Huffington Post made a big deal in the early 2010s about posting articles like “What time is the Super Bowl?” The SEO wars started pretty much immediately after that.
But before that you probably would have had to go to a TV listings site or go to the NBC page for SNL, which may or may not have had the answer, possibly embedded in a Flash interactive.
I remember that era. Honestly it was a bit more work but it wasn’t unpleasant work finding information. We 2000s-era netizens just learned the names of sites like IMDB, wunderground.com, WebMD, Mapquest, ZDNET… different sites for different purposes.
(Just as we had multiple big retailers, in different narrow categories, like CDNOW.com, before Amazon.) All the consolidation into Google for every information need is unhealthy at best.
This is such a nonsense "conspiracy" style take. Google doesn't like SEO. If Google search sucks and Bing or duckduckgo or whatever else rules then everyone would switch over a few weeks. Google has whole teams of people trying to stop SEO from working because if they stopped giving better answers than their competitors then they'd be out of business
> If Google search sucks and Bing or duckduckgo or whatever else rules then everyone would switch over a few weeks
I switched, and every tech-savvy person that I know switched about 5 years to a decade ago. Google stays on business, just like Facebook, because of the huge advantage of being a household name.
This is exaxctly how they built out their dominance in search. Good, high quality answers, no ads. Get the users in the door. Become synonymous with search.
Google absolutely bled cash for their first three years, because they were investing in acquisition. Here we are again.
Then, little by little, you monetise, until the product is once again awful.
They repeated this process with YouTube, for a larger example of "burn cash on user acquisition until vast profits".
We'll see the same with their AI offering. It'll take 10-15 years to the "hot trash" stage, it'll be WILDLY profitable (think of the targetting opportunities!), and then they'll have to pivot to the new hotness, whatever that is.
It's not just death of discovery. This is theft. They scraped your website, built a knowledge base out of it with the promise of sending you people to you. But only now, they use it to answer the questions themselves and cut you off. Classic middleman play.
Yes you are right. However, over the years google search has exacerbated the shady websites that just exist to make a buck by SEO gaming and they even get their content by scraping other sites. Over time they became more popular than legit sites. So chatbots kind of put an end to those kind of sites.
In other words, the search traffic was already amplified by crap and this pulled it back. I am not saying there is no theft because there is massive theft.
No, you can't steal facts. Nobody owns the truth, no matter how hard you worked collecting them on your site.
Plus if anything AIOs link out more to source sites. Google used to give data answers to a lot of questions, which now are being answered with AIOs that cite websites instead. One egregious example I get a lot is sunset times - there's 0 need to cite a website when telling me when sunset will be, a simple python function can do that programmatically.
Intellectual property is a thing. Not everyone is running a Wikipedia. It's just a reductionist take since everything can be collectively labelled as facts.
"who is on SNL" is very clearly an uncopyrightable fact. _Most_ search queries are for uncopyrightable facts, or derivatives thereof (e.g. comparisons, recommendations).
If you write a cookbook, I'm perfectly free to take all your recipies (facts) and publish them in a cookbook of my own. I'm not free to take the artwork and prose (intellectual property) from your cookbook without your permission.
In Nester's Map & Guide Corp. v. Hagstrom Map Co. (1992), the United States District Court for the Eastern District of New York held that trap streets are not copyrightable under the federal law of the United States.
What happens on a football game is fact. How photons hit the field and the players and reflected into the camera is a fact. Now go set up a public real time football game stream and let us know how well "nobody owns the truth" defense works when you get sued.
Video and audio recordings/transmissions are treated differently under the law. You can't compare them easily to written works. Nevertheless, if you watch a TV stream of a game, and then use that to announce facts in real time, that's not an infringement. This is basically the same as what Google is doing with AI overviews; reading sources and extracting facts from them.
I provide a service where you call a phone number, I pick up, and you ask me who's on SNL tonight. I google it, visit the SNL webpage that lists their nightly cast, and I tell you the answer. Then I write the answer down on a post it in case anyone else calls and asks the same question. When I answer that question from then on, is that theft?
> You used to get 10 pages of clickbait that made you scroll down endlessly for the answer.
And this is why, kids, the statement:
"This technology is the worse it will ever be"
is false.
Whether it's search or cars with huge screens or LLMs. Technology can obviously regress, even if the on paper specs progress. It's all about incentives and who the actual main customer is.
That thing exactly turned me off google search years before AI. Bing cloaked by duck duck go and similar privacy driven search engines seemed to be much less cluttered.
Now with AI (I use chatgpt and perplexity) as direct replacement for search engines I find Google as a company even more obsolete than in the past.
So is YT, which is cluttered with ads more than linear TV in American and I thought nothing could be worse… and recommendation engine focused on misinformation more than ever. All interesting content from YT once randomly found is ripped with yt-dlp and stored on NAS to be turned into local streaming service via plex or emby or similar service is my way to access media.
Google created habits of effective discovery of valuable content in the internet and killed it by its own greed.
I never understood the unilateral pre-AI attribution to AI for killing the web when Google search began ignoring search terms and providing irrelevant but thematically similar results years prior to Google force-feeding AI to all users.
The removal of `&&`, `||`, `(`, `)`, `"` having actual schematic meaning on google.com was really the canary in the coal mine. The system ceased to be a tool for information discovery and became one for ensuring the top results were ad driven.
And when you clicked those links, you'd find maybe one wrong answer, and 9 good-enough or maybe even a great answer.
Now, you ask AI if Sonic Advance 2 allows you to get a D rank, it lies to your face that yes. The game doesn't even have letter ranks. Too bad you're streaming on a game show while searching that, letting 8 thousand people see the failure of the modern google in action.
I so much hate google results with these clickbaits, and content farms. That is why I became obsessed with managing search by myself. It is not doable on big scale, but there might be enhancements.
So I have a project with domains, YouTube channels, Github repositories, reddit channels [0]. All have 'ranking', so most important appear first for me.
This project however is already big. 2M of entries. So I have a trimmed version of it, only things I already consider worth maintaining [1].
I also maintain RSS feeds for what is worth [2], and I have other repositories.
I don't like Google. They intentionally broken the search. You cannot say it differently once you understand how broad, rich, and interesting the internet actually is.
Although list of domains does not answer broad questions like "who is on snl tonight", so I still use other search methods.
Well, if you're really interested in the answer, Google AI also gives you the sources, so with any luck by clicking through you'll find a site that's a good resource for the topic, and that site will get some traffic. I do that regularly because I don't really trust Google to give me the correct answer, but I don't know how many others do it...
I hope AI also kills SaaS next to SEO. At least smaller businesses move away from it, larger enterprises will stay hostage longer or it will make more sense in their use case.
But yes, a lot of pipelines are broken right now. If you are a junior dev today, you have a hard time to get into the field. Quite ironic since just a year ago you were too old if you were past the age of 25 for any dev work...
They will probably come up with ‘SEO for AI’ and start injecting ads into their AI answers. That said, commercial web might be dead, however there are still enough people who host their own blog, have strict no-ai rules.
On the other hand decades ago web search was horrible - in most cases what you got was result of full text searches ordered by the rank. You had to go through many websites, read so much information just to extract essential knowledge.
Hyperbolic. In the past, such a query usually landed me on the right info within three or so links, until Google themselves started inserting pages of ads, "affiliated content", spam, info cards, etc. and encouraging tactics which favored tabloids and other high-churn websites. Google killed search.
Hopefully people go back to real social interactions and that kills AI on the web. People just go back to directly contacting businesses rather than exploring their what is now a completely AI generated website. Almost all the models used in the newspaper are now AI.
Yeah, IMO discovery was back when we had web directories, web rings, and stumbleupon. Search IMO killed discovery as you lost the ability to browse categories of sites.
For a good period, Google actually actively tried to combat linkbait, poorly built websites, duplicate content and other SEO shenanigans to try to reduce the art of SEO to "build a good website", even to the point where they published a browser plugin that allowed users to report stuff, but it seems they gave up the fight a long time ago.
I'm not sure why you would claim something that is verifiably wrong?
I just searched "who is on snl tonight?" on the "web" tab of Google (which is Google search without AI) and the first 4 results are in order the SNL instagram, Reddit, Twitter and the NBC website, all of which have the answer easily accessible.
Twitter and the NBC website even have the response directly in their description so I don't need to open the link: "John Mulaney hosts with musical guest Phoebe Bridgers".
Keep in mind, Google search results used to be more decisive and not full of ads/slop. Google's AI answers will see the same fate when Google realizes they need to monetize it & others figure out the Google's AI answers equivalent of SEO.
I already see this when searching for product reviews. "People say this chair is very comfortable." Then you dig into the sources and "people" is a single astroturfed reddit post.
If Google didn’t put those there, people would be going to other LLMs to get their answers quickly. I don’t know how this ends but I strongly doubt any unilateral effort to withhold llm service by Google changes the medium term outcome.
I think it's actually wise of Google to disrupt themselves this way. If they don't the AI companies will do it for them, and they'll have no control over the process nor chance to recover lost ad income. As per the article, Google managed to pull this off without hurting ad revenue (surprise!)
Disclaimer: I wrote the blog post and submitted it
Is the future of providing content hosting markdown files and images surfaced with a sitemap xml file exposed to LLM providers, or perhaps just offering an mcp?
I sure hope so. My AI agent should be my user agent for web access. Some times I want things summarized, other times I want a specific page, and other times I want recommendations for adjacent topics. But in all cases I want my agent to serve information the way I ask for it.
If you don't disrupt yourself someone else will do it on their terms.
The ad revenue will be hurt either way because "providing the answer" is a better product than "providing 10 SEO spam links that probably contain the answer".
Exactly, people treated Google like it was AI long before the rise of LLMs. Now they actually get the answer they're looking for rather than searching through a sea of web links.
I mean, they are going to google search specifically because it isn't an LLM. What is the point of shoving AI there when you already have your Gemini offering as your actual competing product with LLM? Adding AI to the search product just dilutes both product and probably pisses off both the users who come to google search to expect google search, as well as the people comfortable with LLM answers who are getting the inferior "search blurb llm" instead of the actual Gemini offering and walking away with a sour taste in their mouth about google and their LLM capabilities as a result.
Something something old habits die hard, can't teach an old dog new tricks.
I am here on HN on chrome mobile. The omnibox is there right at the top for me to quickly search something. That muscle memory and flow is deeply deeply ingrained even if Claude or Gemini is the adjacent app open
1. People have forever typed questions into Google. Why open a website when it's already your homepage?
2. I honestly dislike how every single chat I start appears in the sidebar. I only want to ask one question, I don't want to see it every day for a week. Google is decent at that (even if IMO overall pretty stupid compared to free Claude)
When I first connected to the internet in the mid 90s the www was young, and there were still traces of pre-www internet information services such as Compuserve, AOL, etc. They acted as subscription-based computer networks that provided access to curated content (and more). I never got their point back then, but I understand they were walled gardens of information.
Fast forward to today, we have witnessed the open web grow, provide capabilities that were unthinkable of, but information-wise, it looks as though we are going back in time. AI answers/overviews shield users from the complexity and uneven quality of the raw internet, but at the expense of cannibalizing traffic to the origin, thus becoming the information gatekeeper and ending up owning the discourse.
Knowing that laymen use frontier chat providers as search engines too, all in all these are becoming centralized information brokers that filter the wider internet to keep users inside a single platform.
Although this is true, before I would scour the internet to extract the needed information, with an adblocker, without mentioning the sites to my friends or engaging in the communities. Now the AI search does that for me. For the communities that I do engage with, e.g. some open source dev activity (github+forum) still needs me to manually engage with the website. So AI replaces the leeching part and automates it. The human engagement part AI can't get in between, and I don't see chat changing.
You will be in for a ride when they come back in much more subtle and harder to avoid form. I just hope some providers will emerge that offer the ad free experience in exchange for money. Sadly the social media age was lacking that largely.
What's your opinion on this take of mine, from another comment of mine on this post:
> "Ads will stay but change. Companies still have good products and services to offer, and can't rely solely on word of mouth. Websites will still offer related products, such as a Linux site having ads for computer products. Users searching for products using their agents will then be served these ads, but only when relevant. My agent looks up how to close vim, it will find the answer, plus a few ads of automatic vim closers as paid npm modules. The agent will then tell me: you can close vim this way, or you can use commercial offering A ($XX /m) or B."
The problem is perverse incentives whereby if the LLM provider tracks attribution/conversions (which they likely will,) then then inventive is for the agent to suggest the paid version over the free one. Just now its not a square box on a page to refute, its weights that are unseen.
I believe the only solution to this is open weight models hosted locally (or on your own rented hardware/api). Closed models will be incentivized to profit max especially once the honeymoon of subsidized subscriptions is over.
Makes sense. What we might see is that more expensive providers pop up that are less biased, with the cheaper ones running partly or wholly on ad money. Of course I'd prefer local if cost and quality loss (if any) is tolerable.
I know that I reduced my own search traffic by way more than 40% with AI. Because what I actually want is answers, not searching the internet. And right now, AI is the one that gives me the answers as fast as possible, within my tolerance for falsehoods.
Same way that internet searches supplanted me going to the library, or opening up an encyclopedia at home.
Almost all replies are false, cut information, are outdated or are not relevant. Though their interface for me did not change.
And yes, I am aware that rewriting the same query multiple times reveals more references of relevance.
To me it is a absolute regression.
I can't understand how you can be so happy about the state of search.
I need to clarify I'm not happy about Google's AI summary, I'm happy about my ChatGPT Plus subscription :) In High thinking mode I get very good answers to whatever I seek.
For search, I use duckduckgo btw. UI seems like if Google UI frozen in time, from back when it was most useful. Results... sometimes very dumb, but 99% of the time I don't redo the query in another search engine.
Myself, and clearly many others, find it acceptable. I rely on it for trivial information that I know I can trust LLMs on. Just the other day, I was curious about when some senator was in office, it’s highly reliable for these trivialities. And not having to sift through endless ads/clickbait/spam? Sign me up threefold.
And btw, when I want less trivial information, I go to Claude (Or other LLM of choice). It’s a great tool for collecting and synthesizing information in the way I want to digest it. Traditional “search” is dead, as far as efficiency is concerned.
Yes, people want answers. But what people need is to work a bit for those answers so they actually have some idea of how reliable they are. If you just get answers you gradually lose the ability to evaluate how believable those answers are. LLMs are comfortable because they produce in you the illusion of learning things fast. But the job of verifying sources and reasoning about the information you get, is just as time consuming. But because the answers seem so polished you won't. Which makes your understanding less robust over time as you grow lazy and complacent.
Also, it is unfair to judge search engines on what Google offers today. There was a time when Google cared about search quality, integrity of result sets and you didn't get a face full of ads with perhaps one "organic"-looking link somewhere on the page.
Google has become ad search. Not web search. And we should be comparing LLMs to actual search engines.
>Which makes your understanding less robust over time as you grow lazy and complacent.
I really wonder about this. Of course it has been bothering me as well. However, when I was young, I have been listening to the same argument wrt/ calculators and the internet vs the good old ways like books and the library. I think this will be the same somehow.
Verification is very easy with AI, I just tell it to bring sources when I feel like it.
The obvious end to this is ads in ai output. The interesting question is, will google be doing it or will the new (or probably current seo) be writing pages specifically designed to be added to steer ai answers?
As the visitation to other parts of the web no longer served by the search model drops and pages are not updated, it will be easier to game the results and steer answers for some things.
X is such an idiotic name that if they put X in their sentence we wouldn’t even know what they are talking about necessarily because X is a universal identifier for any variable.
>Google could share revenue with authors, like X does
Google shows a directory of results related to a query. Research + result selection was up to the user. For some use cases, this is a BENEFIT (academics, very specific searches). However, AI can also handle that, if people want to use it that way.
For MOST use cases, AI is just better. I want to cast a wide net and be able to direct that query to other things. "Who was in the movie I just watched. > What other movies was she in > Is she religious > Are Mormons really not supposed to drink coffee?"
The surfacing of the information I'm looking for is faster, the flow of information feels more natural, and there's no learning curve. Just ask a normal question and get a single (probably correct) response that pertains directly to the question.
And most importantly, the vast majority of users don't actually care if the information they're reading is correct. They just want an answer, if there's any amount of truth to the response, it's 'good enough'.
You’re describing search as if only the interface matters. As a searcher, I want there to be a steady source of good information today, tomorrow and a decade from now. A great interface that leads to less information being published is not a good thing. Short term, sure, but long term, no, the power of being able to search the web has always been the diverse variety of information. Instant answers are only going to get worse as the incentives to publish information dry up.
Moving search to the chat interface has some other advantages, like better control on the context and output. Paradoxally it took the AI turn for Google to understand e.g. that I want query based language detection. Or whether I want fuzzy search or very specific things.
But that's beside the point IMHO. We could have had AI _and_ search, combined in a cleaner way. We wanted search to improve for so long, it didn't have to be throwing the baby with the bath water.
> there's no learning curve.
This really drives home how search is different for everyone. It took me a while to find ways to get the results I want the way I want.
As someone who did a fair amount of SEO in the past - good.
Search sucked so bad for the last decade and maybe AI can actually sift through all garbage better than whatever-nothing google's pagerank tried to do.
The only danger right now is that people will gate information or stop creating. I think we desperately need cheap easy blogging networks like reddit where people can journal their findings and opinions without restriction because people always want to share, even without being rewarded.
Whilst I agree that search sucked, I am not sure if this is a *good* development.
> The only danger right now is that people will gate information or stop creating.
To be explicit, some people create for the joy of it. At the end of the day, I personally am not harmed by people reading a condensed version of my writing through an LLM, because I don't make money off of it. Websites that rely on actually being visited and read do though.
Google is basically monopolizing the ad business on information right now. Not that they didn't attempt to do it before, but this time, they seem to be way more successful.
How do you prevent people gaming your search algo? It's not that Google invented SEO. It's people trying to reach the top. That's not even the problem. The problem is, that they achieve that with scam sites. Greedy people who flooded the web with garbage killed search.
At a technical level I basically agree, it's really hard to have good search when there's an entire industry built around trying to game the signals you use.
However I think at a system level you could place some blame with google, in the sense that they basically built the whole system where ranking well in search is something that is profitable for people to do, because they control the search and they control the ads that people put on their SEOd websites to actually make money from clicks.
Even if Google didn’t have an ad business and they had to use an ad competitor, do you think that would stop people from trying to game SEO or do all the other things you mentioned? (I think people would do it regardless of whether Google sold ads.)
It's true our search traffic took a dip, but our sign up rate and conversions are significantly higher. AI feeds in better leads than just a single search query. The users have a chance to ask AI back and fourth about market options, and by the time it leads them to our site, they're a very qualified lead.
> Earlier this year, a small site called HouseFresh, which is dedicated to reviewing air purifiers, published a blog post that really crystallized what was happening with Google and these smaller sites. HouseFresh managing editor Gisele Navarro titled the post “How Google is killing independent sites like ours,” and she had receipts.
I can’t speak to the quality of HouseFresh (I’ve never been in the market for air purifiers), but 99.9% of these “review blog” sites are low quality and exist not to help readers find good products, but to gain referral revenue through clicks.
So for these sites, I’m not sorry that “Google is killing them” (and I’d argue it’s LLMs that are killing them, not Google singlehandedly).
I use to use these types of sites all the time. Not because I trusted their recommendations, but to even learn what the popular products in a category are.
How I just tell an llm what I am interested in buying and it gives me recommendations.
Their methodology seems ok, but clicking through their pages and looking around, their comparison method is all over the place.
* In many posts they lament if a device doesn't have an app with it (???).
* They have reviewed only 2 Honeywell brand filters. Even though they are in the same ballpark of cost, area, size, and performance. Why not more of this brand? And I had to use google to search their site.
* If I search their site for "Honeywell", I see one smaller unit, not the other unit they reviewed, and then at the very top "how to fix the top problems with Honeywell air purifiers".
* How are they able to provide such expert advice on how to fix Honeywell products, if they don't seem to review them? Unsure, if they didn't remove the Honeywell review pages I'd have more confidence
* They admit on that same FAQ page Honeywell is a huge manufacturer. Where are they on their site, and why does it seem all those pages were removed? I checked waybackmachine, and those pages used to exist. Maybe they did a sitewide purge of other brands?
* The latest I could find for any article referencing Honeywell was in Dec 2025, when the purge of Honeywell review pages began.
Yes, I have Honeywell purifiers which is why I picked that brand to search on their site, as I have familiarity. They represented the capabilities of the purifier accurately.
But it seems to have near identical, in some areas better, performance than the ones getting recommended on their site.
I tried the same kind of search on a niche I'm more familiar with (mirrorless cameras), and the LLM results are pretty mid compared to the tradional results.
I would have expected a site like DPReview to rank pretty high, but it's only the 5th LLM result, which is about the bottom of the 2nd scroll page on my screen.
Of course it's the first in the "organic" result (way way below in the page, about the 4th screen of scroll down to me)
But the AI has just been trained on these sites (or uses search to find them) so you are just getting recommended the same product that would have paid these sites the most commission...
Chatgpt has reduced my use of Google search by a lot more than 40%.
In the long run will there be room in the market for Google's (free, presumably low-spec) AI search results? Paid services give you models that provide much better answers.
Maybe? Hard to tell who the winners and losers will be. I think it’s a better gamble to give away free low-cost AI today than to continue to offer the Google search results of 2023.
I wonder if eventually Google will stop serving search results entirely and just become a front for their AI. After all, why give people results they won't check? Why focus on an outdated product? And if you're a big co with a bias, it makes it easier to control the narrative when you're front is the only source for answers.
If this happens, I would guess that Google will probably do this within the next 10 years, maybe even within 5.
I wonder what would happen to websites. I wonder if someone will step up to fill Google's niche? I use DuckDuckGo, Ecosia and Kagi off-and-on but will there be some critical-mass moment where everyone switches to get away from AI slop?
It seems unlikely this would happen. What is likely to happen is the opposite: people stop using DDG etc because they offer no AI summaries; you have to sift through a lot of websites to get the information you need instead of getting it upfront.
For my common searches, I find that DDG has few of the problems people use to complain about Google. And, across all kinds of questions, it certainly does have 'Search assist'. It gets out of the way of better sources... and abbreviates the more obvious ones.
Websites will be a niche, just as apps. The future is a direct, agent-accesible API such as MCP. The MCP service will be announced for discoverability to search engines, and is enriched with metadata that describe the way the interface can be used.
Human interfaces, such as websites, will be created by your local agent in whatever way needed, and will read/write data through that MCP interface. Ads have no future in the web of tomorrow. Probably micropayments.
Think of "Hey agent, create me the soccer news, format like the NY times of the 60s, and print it on my printer in the office" instead of opening your favorite sports website.
Ads will stay but change. Companies still have good products and services to offer, and can't rely solely on word of mouth. Websites will still offer related products, such as a Linux site having ads for computer products. Users searching for products using their agents will then be served these ads, but only when relevant. My agent looks up how to close vim, it will find the answer, plus a few ads of automatic vim closers as paid npm modules. The agent will then tell me: you can close vim this way, or you can use commercial offering A ($XX /m) or B.
I would challenge that. Because I see in the midterm future that I run open-weight local agents - meaning, I am in control and they ultimately follow my command over all else. And inevitably, I will have "remove all ads, product placements and upselling" in my global AGENT.MD file.
Because of that, companies will try to prevent becoming a mere MCP data supplier and force pre-made frontends down our throats. The question is how good will my local agent be able to crawl this - there are already website-to-MCP providers that make any website available via MCP, such as [0] which is YC backed.
My traffic is down over 80%. The information I create is really easy to steal, but it's as expensive as ever to create.
I have built a lot of infrastructure to surface that information. I've set up group chats, built relationships, and created tools. Now all that work just feeds the AI that is putting me out of work.
By December next year I'll probably have to put the website in maintenance mode and find another job. The problem is that AI also made software development absolutely fucking miserable. At least I have savings, but if this whole AI thing fails it will also tank the economy.
These companies are very, very big. Since they control all the roads to the independent web, they set the terms. You can't hurt them without hurting yourself far more.
I would argue that this value capture was the goal all along. I saw it coming from a mile away, and I expect it to get far worse.
This is sad. Your portal is a work of art! I wonder if you are throwing out the baby with the bathwater.
Your content is amazing and you're just putting it out there for everyone to see. I hate to say it, but if I were you I would go the Patreon, etc route and have some basic information on your free tier - enough to keep some traffic - and work with paying customers. This might require some changes in the way you work but I feel you are already well-connected and liked and have built a healthy community. I think you will get 80% of the benefit for the community and you get to keep doing what you love.
You'll be the Berlin-guy and accessing your information in full will only be $1,99 for the "Pro" tier and $4,99/month for the "Deluxe". I'm only half-joking. I think it's time for people to start putting a premium on everything especially their time because the corps ain't giving freebies and why make them rich?
Absolutely not. One of my guiding principles is that this information should be free and widely available. It should reach those who stand to benefit the most from it. It was easy to uphold that principle when the going was good, but now I get to put a price on it.
I have helped Berlin build an official equivalent and it was dead on arrival. The federal government has also consulted me and offered to hire me for my advice a few months ago, but I haven't heard back since.
The government cannot build, buy or maintain anything like my website. It's not that I am so incredibly awesome that they can't match my work, but they operate under a very different set of constraints. They'd rather burn a million euros on a dud than bankroll one guy to do whatever seems to already work.
I have many helpful contacts in that area, and they all seem to agree that the government/grant approach is not worth it.
On the other hand, I would love to work for them if I am forced to get a job again. Some parts of the administration are really cool: CityLab and Digital Service for example.
What if you open Patreon with subscriptions, the guides will be free, but some users might pay you for continue creating them. You can add some other tiers like personal recommendations. You already have some users/community so you can start exploring options with them.
There's a /donate page. It's getting €60 a month, down from €200-400 at its peak. My birthday and the site's birthday are almost the same, so I buy myself something nice with the money every year. It's how I bought my favourite item in the world: my bike.
You are damned if you do and damned if you don't. Hard paywall will likely cause additional drop in traffic from google search. And, even with rate limiting, someone will but a single account only to rip all data from site.
yup, it's a slow death of small content creators; small businesses which heavy related on organic traffic (putting our hand crafted articles, knowledge cases etc.) are also struggling - the trend is clear; it's interesting that a lot of HN crowd seems to be happy with this; I wonder from where you will get all your answers when there will be no new content created?
A big portion of HN readers/commenters are absolute idiots who seem incapable of thinking at all. I think this environment brings a lot of people who, because they are semi-technical, think that correlates to being smart. Reality is, some of the least intelligent people I have ever met, have been through my tech job.
I also think that being really good at a technical thing doesn't make you smart as much as it makes you attached to and have a vested interest in what you're doing. One still has to do a lot of introspective and philosophical thinking to be generally smart; I've met many narrow-range-smart technical people who were absolutely stupid outside their domain.
I think that your experience just highlights a problem that 20 years of abusive ads masked - we need a proper payment solution for the web. Right now everyone is chasing the big 15$ subscription for his substack, news site or premium service.
If serving me ads brings you 100 microeuros when I read a page on your side, i don't mind paying this. The problem is that there is just no service with which I could send them to you. Same is with news, forums, reddit or whatever.
If google (and other search engines) don't drive significant traffic to websites, then what's the point of allowing them to index your site. RIP the open web as we knew it
Though actually, it’s becoming more and more common to slap cloudflare bot protection on your site, soon the LLMs won’t have any access to current information.
If you want to win the war on AI search you just have to have a browser and an LLM answering all search queries. I don't really care who's behind the answer as long as it is right (which 99% of the time it is) so there is no loyalty at all, no need to switch, no need to try the competition, nothing. AI is the answer and it will get better with time
I use Brave, I don't know if it uses Leo (brave search) or gemini, or chatgpt, I just know the answers it gives me are the ones I am looking for and most of the time I extend the conversation with some of the options offered, hardly click on any link below the answer provided (which most of the time those links are garbage)
Who will feed the AI if there is zero reward? Someone has to review vacuum cleaners and try local cafés. Some lawyer or mechanic needs to chime in based on their experience.
Who will do that if it all gets stolen and stripped of any attribution?
You could argue there's now more niche (coding) knowledge than ever before. Do we seriously think that "de-identified" enterprise data isn't making into the training sets of LLMs? None of that proprietary code would had ever been posted to SO, but it's being posted millions of times a day to Claude and ChatGPT.
> Someone has to review vacuum cleaners and try local cafés. Some lawyer or mechanic needs to chime in based on their experience.
Youtube, google maps, reddit are far superior for these purposes IMO. I don't remember the last time I actually went on any independent blog/website to read a review on anything. I've always found them to be full of SEO crap and not super useful.
Are we losing valuable traffic? The person who wants the quick answer seems like a less valuable customer than the one that knows they want to read a full article.
Google's AI solved a fundamental flaw with old search engines. SEO broke the internet over and over. AI... LLMs being used for searching information is really a good thing.
It's called GEO and the level of the craft in that field is currently akin a bit to astrology. Same as I guess as the regular sharing of prompting advice.
I used to work on web search engines. I worked on several search engines. Google's among them. I was fond of web search engines.
Google has not been a very good search engine for a long while.
It has become the opposite of what it once set out to be. It set out to be a search engine were relevance was king and where ads were curtailed. You got just a couple of ads, but it would more likely be relevant. They were also clearly delineated as ads. This made both search result and ads valuable. It was clean, it worked.
Then gradually people who don't stand for anything seemed to take over at Google and things god muddled. The search engine grew a little worse every year. Until it was utter shit.
Earlier this year I mostly stopped using it because it no longer delivered sufficient value. What I started using instead was a poor replacement for a good search engine, but it was better than the mess that is Google search.
The AI answers to searches are not search results. They are something synthesized from what it thinks I am looking for. It is a poor replacement for actual search.
If I'm going to ask an LLM I'd rather ask a higher quality LLM. Not whatever that thing Google has plugged into their search page is.
OF COURSE, search traffic goes down when you gradually degrade a product and then try to somehow make it relevant anyway by slapping a slop-generator on it. This is not a surprise.
And of course Google search is not coming back. Google is not about making products people want. Google is about making products they think they can monetize. I happen to think they are wrong. Perhaps other people think so to. But Google want to go in this direction so they will defend it even though it may not be the smartest thing to do. Careers, personal prestige and pride is riding on this. Of course they are not going to listen to informed consumers. That would be losing face and admitting they are flailing.
Google is great at fumbling products. The turn has now come to fumbling search.
This is a great opportunity for someone to take search away from them. Because search still has value.
Google is reducing search traffic by other means too. e.g. Firefox users, VPN users, ad blocker users, etc. are now often required to solve a captcha and are sometimes even prompted to login after that.
The time has never been better to switch to duckduckgo.
I think people have to make a tradeoff between trust and effort here. Obviously, Google's AI answers give you less effort but can you trust every summary? Earlier, I used to see that Google used multiple sources but references were provided from websites you wouldn't normally visit. Over time, this has become better and an internal ranking algorithm is probably getting the most clicked/trusted websites as the source of information rather than clickbait ones.
For most, effort matters more than trust and the stats prove it.
For me I use Google maybe 99 percent less I just use chatgpt for information. Not everything but definitely in the high 90s. YouTube is different but there i am looking for entertainment or experts I trust
I’m not a fan of the false equivocation going on in the comments, because it gives the impression the poster is ‘in the bag’ for AI, like this, the top comment as I write this:
> You used to get 10 pages of clickbait that made you scroll down endlessly for the answer. The best product for most questions is to just give you the answer.
(I used to have to go to the bottom of page 1, perhaps page 2 to find the answer. It never took me 10 pages, even in the worst instance but even taking the OP at their word…)
The false equivocation here is the idea that Google’s purpose was to ‘give you the answer’. For the 20+ years prior to Generative AI, Google’s Purpose was to surface the web page with the answer.
Google did not have the answer. Google showed you where to get it.
Generative AI both turned this on its head and obfuscated the answer (due to hallucinations).
If GenerativeAI had simply given a 100% correct answer, that’d be one problem, but it does two things:
1. It gives an answer that could very likely be incorrect (due to hallucinations, or due to how it strings together tokens to form those answers)
2. It doesn’t always source the answer it gives.
The web, as it has worked for its history until GenerativeAI, was the fact that links to other sites were how everything worked. Google was the means to find those sites, and now Google has changed its relationship with the rest of the internet.
This is wholly destructive, and not in one of those good ways. If you’re ad-based (Which the entire internet is built on), you’re going through a rough time. Bandwidth isn’t free, serving content is not free, and your time certainly isn’t free. In an attempt to move from “This is how you find information” to “this is the information”, Google has risked destroying its own model and the advent of the internet with it.
We can be irritated by SEO and the content farms that came before; but to equivocate the advent of AI with them is absolutely missing the entire problem that Generative AI causes: there is no more business model for the internet as we grew up with it. Now, websites are incentivized to hide as much information as possible behind a paywall, and perhaps this will end up being a good thing, but it will certainly be the death of the open and free internet.
It would be nice to see a website that classified websites well enough to reduce each site down to ONE link per search result page, tied to the result websites' main themes rather than some subpages where blogs lurk, perhaps using AI to sort results but not replace them, instead of offering a very suspicious summary and lazily offering anything that looked reasonable.
I recall old AskJeeves days (which became Ask.com and which closed May 1 this year), it restricted links quite tightly, giving few results. But Google came along and gave tons of "close" results, giving users the impression that it had more than it did even though most of it was and always has been irrelevant trash.
In a way, we've gone full circle: from few answers to a boom of answers to burnout and back to summary answers. Now ChatGPT has been eating up Google, but Google's business model has evolved, whereas AskJeeves did not. Maybe if Ask.com still had the capital, they might make a good AI search results co.
Accept the reality that the AI answers are for many searches an improvement for the user. Would not be maybe in the web of 20 years ago. Now it was a shit party of useless sites, advertising of all the kind, and other garbage.
Absolutely, we finally don't have to scroll through the flufftext and ads, click cookiebanners for 200 trackers, and skim the sponsored bullshit when we just want a tortilla recipe.
I haven't used Google Search in months.
Two main reasons.
(1) Every time I login Google Drive I get three 3 exclamation marked warnings that my drive it almost full (its 94%) and that I better upgrade to the paid plan (I wont).
(2) I have several sites and that it takes now several months to get them fully indexed is hard to stomach for a company running a search engine.
probably to indicate that google has become a source of tiny annoyances without nice solutions besides paying for an extended plan (to save space on google services / devices you need extra apps that are not very advanced either)
I don't know, it sounds like a very reasonable UI design? The alternatives:
1) Google somehow provides infinite amount of storage for free users.
2) Google doesn't make it very clear that you're about to run out of storage.
Google already has a generous plan for free users (15GB vs iCloud's 5GB vs Dropbox's 2GB). It also has a simple view to show you which files are occupying the most space. I genuinely don't know how this is a bad thing.
Hmm maybe I'm wrong but the file cleaning view felt too limited and didn't help me clean things hence my opinion that it's a marketing nag disguised as information.
There would be no complains if Android had easy way to point all backups to other service, for example to OneDrive. As disliked as it is I find it significantly cheaper. So, complaining about those scary Google drive messages is justified - they want you to pay more for already overpriced service.
Google has nerfed search so much that they need to find other revenue grows stories but they destroy the original USP and we are now back to the time before gmail.
Lmao. Google makes more revenue from ads than all manufacturers make from disk drives combined. That's the scale of their ad business. They don't need to sell storage to some consumers for "revenue grow stories."
It’s a broad representation of US search, not a spam-focused list. But it does have some biases.
They source it via Google Ads and Bing Ads services. In practice, they grow the list by asking Google’s Keyword Planner for related terms and collecting things like “related searches” from results pages. But sourcing keyword queries like this tends to skew to queries with commercial value, reflecting what advertisers care about.
I wonder what's going to happen to Google ad auctions, and the how sites track and optimize their ad spend? (I wrote a paper on this in grad school, but Google's current AI first page now seriously breaks everything)
Over the last couple of weeks I've been having to communicate with my real estate agent a fair bit about arranging a new place to move into. As I'm annoyingly bad with memorising numbers, each time I'd call I'd pop in the name of the agency into google to hit the phone call button.
I'd put in the exact agency name and suburb name.
And every single time, google will return several results of other totally different agencies, then an advert for some other random thing, THEN show me the result I actually asked for several scroll events further down the page.
Even without AI, their search results are just bullshit. With AI, it's even worse.
Anyone remember MySpace and why it lost to TheFacebook? While there are a few factors, people prefer a uniform interface (and a low 'latency') when asking questions. For myself, maybe I'm old, but I don't care much for the variations between websites anymore when I'm searching for something.
Instead of building content ecosystems and partnerships with content creators these big LLM players now are destroying the very thing that feeds their very business model and technology: original content.
Imagine if Netflix started stripping down the credits and titles from all the movies on their platform only for the movie industry to crumble and stop producing any movies. How would that work out for them in the long run?
Is it Claud or ChatGPT that are going to write every piece of new content for these platforms to keep feeding on, because for actual content creators what is the incentive now?
This is true but top recommendations still come from search results only. As legitimate organic content becomes more scarce, recommendations will be random and biased to whoever is gaming the system.
AI search is now the default thing anyone is doing I feel. We are searching directly on ChatGPT, Google's AI mode and others for getting the result directly.
Google needs to find a new way to distribute its revenue to the sites it has crawled, or the entire ecosystem that Google once helped cultivate and thrive will collapse.
AI-generated slop sites are a big problem which drive me to use AI to answer my question. Legitimate sites with good info get buried under the slop so regular search is no longer good for me to find info.
Maybe it's just me but I click on the links in the answers. It's pretty disconcerting when half the references don't actually support the answer, or are some reddit post with one response who is just talking out of their ass
Right now, everyone says it's a good thing. You have a question, you get your answer. Done!
The problem is when you don't visit the website where the answer came from, they start believing that no one cares about that information, and stop contributing. Give it enough time and the only answers we get will be from tweets. And they will be as reliable as you can imagine.
That was my guess too, but it seems like Google pulled it off so far. Clicks on ads didn't change, so Google is taking care to not cannibalize their own revenue.
Did they do unsustainable things that hurt their advertisers in order to temporarily goose ad clicks and earnings? Almost certainly, and I've heard anecdotes about that.
I have one of those anecdotes: A few weeks ago i tried to do a display campaign, it was like 90% scammy sites that pop up in mobile games, and you could exclude hundreds and hundred more will appear. Before that I was running some search campaigns and i had to configure each keyword to be exact match because my ads where showing everywhere and exhausting my budget. Also I had the impression that my competitors where clicking my ads to ruin my budget.
It was heartbreaking because it was such a good service like 5 years ago.
kagi went free the other day , you may wish to unadapt and undie ,because this particular search engine seems to cut thru SEO quite well, havent used it much but it feels like using google from 15 years ago
couldnt use it on my android shitbox either, however i found the couple second delay on pc to be worth the wait for actual search results , it feels like the first time in ten plus years i can find what im looking for
Many sites absolutely deserve it, unfortunately. What a blessing this is to get the information you want at a glance. Think of searching for recipes in the pre-AI era. You had to wade through paragraphs of organic and hand-written SEO slop. Good riddance.
the worst part is the ad revenue they have lost on search being more efficient and pinpointed. Seems youtube is taking up the slack by over compensating with ads. sometimes 7 ads back to back in one go!
But nothing good lasts, marketing in ai services is now opening and we’ll AI can really rewrite custom campaigns per targeted user, before it was obvious now… not sure what can be an ad!
But yeah discovery is good for some things but fact based stuff or gossip a quick ai answer is perfect
Google web search's goal has always been to get answers in front of eyeballs. It has never been to drive traffic to sites. Ads and Search are different product areas. But nobody believes these things.
A little discussed aspect of this isn’t just the AI answers taking up 30-40% of the real estate. It’s the ad slots taking up the other 70%.
If you try a Google search now without an ad blocker it’s totally insane. You don’t get the actual results at all above the fold anymore, and on smaller screens it’s sometimes 2 full scrolls before the actual results.
It’s a monopoly abusing their default status in browsers (who they paid to be the default) to manipulate the entire digital economy to favor companies that spend money on Google ads.
I couldn’t be happier that ChatGPT is killing their business, after monopolizing email, online video and the open web to extract absurd rents off the original vision of the internet (which they ruined) they deserve to die.
It's not that. People have purchased autocomplete traffic and filled all search results with fake generated things. Since TPU's are at 16,000 tokens per second, by the time you hit enter and the page loads, every single site is fake.
You can try any search on any subject and every single site generated is fake.
Somehow, they have also managed to get copies of some emails, and impersonate the sender and send fake emails too.
Here is an example of a real email I got from a real sender:
Hello Robert,
While reviewing engineers building practical AI-powered products, your recent work developing and operating public AI services stood out, particularly your deployment of a quantized Qwen model on CPU infrastructure and your work coordinating AI coding agents to build, test, and deploy functioning applications.
I'm (name), an Independent Career Advisor Consultant. I'm reaching out regarding a specialist mandate focused on AI product engineering and applied agent systems, where your experience building AI-enabled applications and your expertise in local model deployment, Qwen, agent-assisted software development, browser applications, operational controls, and production deployment appear particularly relevant to the expertise being sought.
I'd be pleased to share a brief overview or relevant listing for your review. Should the mandate align with your interests, I can guide you in positioning your experience against its requirements and developing tailored application materials before you move forward.
Would you be open to reviewing the details?
Best regards,
(name)
It's meticulously and very carefully written, after a complete review of my technical portfolio. The person who wrote it looked through my Github and publications, and is a senior hiring manager level person.
However, when I replied saying yes, I received this impersonation sent shortly after my message. As you can see it's obviously completely AI-generated and completely fake.
"
Hello Robert,
Following your interest, I have identified four opportunities that align with your experience in AI-powered application development, autonomous systems, local model deployment and AI assisted software engineering.
Here are the positions for your consideration:
1. Senior Software Engineer – AI Agents
Company: Traba
Location: New York / San Francisco, USA
Work arrangement: Not specified in the posting
Expected salary: $200,000–$240,000 annually, plus equity.
2. Senior AI Engineer – Multi-Agent Frameworks
Company: ClickUp
Location: United States
Work arrangement: Fully Remote
Expected salary: $200,000–$250,000 annually, plus equity.
3. Senior AI Engineer – Core
Company: Hilbert's AI
Location: San Francisco, California
Work arrangement: Hybrid
Expected salary: $180,000–$215,000 annually, plus potential bonus and equity.
4. AI Forward Deployed Engineer
Company: Regal
Location: New York, USA
Work arrangement: Hybrid
Expected salary: $135,000–$180,000 annually, plus equity and potential variable compensation.
I have prioritised opportunities involving practical AI implementation, agent-based systems, software development and the deployment of AI-powered products.
Which one or two of these opportunities would you be most interested in pursuing?
Once you identify your preferred positions, I can help you review their requirements and prepare your application materials accordingly.
"
It looks like a human scammer might have lightly edited this text, which is obviously completely AI-generated.
So not only has AI filled the entire Internet with total useless crap, scammers have managed to buy access to private emails and redirect real inquiries to their worthless scamming.
AI Overview is an attempt by Google to re-do the older
private web (AMP) again, just with a different focus,
e. g. redirect all traffic onto Google, via its AI
slop summaries. Many people noticed how low in quality
these overviews are. I guess for some smartphone users,
though, being able to access whatever comes first, may
be useful; for a desktop user I don't see any benefit
but most assuredly many disadvantages, which were already
pointed out by many folks (also on youtube which ... is
also owned by Google).
Yes this sucks for the open web but anecdotally we've been getting new clients cold off of, apparently, chatgpt. I guess it ingested info about our co-op and has been recommending it to people that ask it for agencies? I would love to know what made us attractive to the LLMs so I could advise clients on same, but I have no idea! My best guess is just having lots of FOSS projects and generally trying to make useful blog posts rather than SEO slop. So... basically the same as the old "actually good" SEO advice.
Looking at the analytics of my product (memory for AI agents), more and more traffic is starting to come from AI assistants: ChatGPT cites us, suggests our plugin, and users come. Same with Claude.
ChatGPT already runs ads in its answers. Haven’t tried it yet, but it looks interesting.
Google has already put AI into search, and I think soon it will integrate deeper and deeper with Gemini.
Looks like everything is heading toward doing content, services and purchases all through agents, without visiting other sites.
> What's the point of writing blogs or travel guides anymore, for example?
To share your love of a place or your thoughts.
I never thought monetizing these things was a good idea.
I've been writing my blog since like, 2010, never served an ad or made a penny off it. Got a job once through it though, so I guess that could be a reason if you really needed one?
It's not all about the money. My writing is a bid for connection, a search query for like-minded people. Cyclists still write to me about my Singapore border crossing post. Why would I write anything if my audience was only bots?
I'd give my neighbours eggs from my coop. It's how you turn a neighbourhood into a community. I would not give my eggs to a Best Western breakfast buffet two towns over.
what's the logical terminus of this? Google doesn't have feet on the ground, real world reviews, reports, testing of products, material ground truth comes from people who rely on traffic. This is self-defeating in the end because Google won't have anyone to syphon data from.
It's basically what happened to local news when it got eaten by social media, now you have no real world reporting any more, you just have some Simulation & Simulacra world where people and bots yell into the void on twitter
I'm somewhat conflicted about AI overviews. On one hand, as other comments mention, if Google didn't disrupt themselves, others would. Also, and this is just a personal anecdote (but it jives with some of the data in this article), but the primary times that AI overviews keep me from clicking through is when I just want an answer with information (and sometimes I will click through if it's to a site that looks high quality). So it would appear the sites that get hurt the most are content spam or at least "content spammy" sites that try to draw you in with usually mediocre content to buy something else or show you ads.
But at that the same time, Google is now just vacuuming up all the revenue that used to go to other sites, and that feels blatantly unfair to me (remember back in the day when Larry Page said "We want to get you out of Google and to the right place as fast as possible"? My how times have changed). This is one reason why there is so much general distrust of these tech giants: they start out all "Hey, look how mutually beneficial this all is" and then ends up at "I am altering the deal. Pray I don't alter it any further."
Does this mean we will finally see the end of Google as an all-powerful company that devours personal information? I have never seen a company that lies so much — with that advertising slogan, "Don't be evil", which looks so false in hindsight — that exploits our personal data so heavily, and that is so cozy with the US government at every level. It is terrifying to think, for a non-American but for an American too, how much personal knowledge this company has about us... and of all the agencies and public bodies that can access that information with nothing more than a court order, or sometimes not even that.
Look at how it has polluted the Internet with countless bland ads, tracking metrics, anti-adblockers and so on.
I can't deny I'm eager to see it shrunk down to nothing, just like Yahoo.
This is probably a good thing. At the end of the day reducing SEO slop and encouraging real answers is an outcome that is absolutely possible through AI integration
When you typed who is on snl tonight? You used to get 10 pages of clickbait that made you scroll down endlessly for the answer. The best product for most questions is to just give you the answer. That said, the death of discovery is worth mourning. I hope someone solves it but this was dying before ai came along. SEO killed the web, AI killed SEO, and so it goes…
The big thing that scares me with this is how it's driving further traffic consolidation to the big players.
It feels like the web has just become either social media or AI. There isn't any room for smaller sites to find users anymore.
The "web" is basically 10 websites and 20 apps at this point.
I want to argue against you, but I just can't. Beyond the hyperbole, maybe we are on the cusp of change. Stuff being posted is AI slop, but maybe it is the Geo Cities obnoxious site of 1999, when an person does it because they want it and not a content farm.
I moved to America as a kid, in the 90s.
Obviously, I'm not thinking about "efficiency" and "productivity gains" here, but I do have fond memories of ordering Chinese food over the phone. We had a few places and their phone numbers on the fridge, like that takeout place, or places like Pizza Hut (the dine-in experience was awesome!), and so on.
It felt like places had soul, even the Taco Bell or Pizza hut; now everything is corpo-slop or ghost kitchens.
I'm far happier making my own food at home, and I have zero interest in "supporting" local establishments that want to shake me down because they are run like PE sweat shops.
we have snow crash at home!
> It felt like places had soul, even the Taco Bell or Pizza hut; now everything is corpo-slop or ghost kitchens.
Yeah, I don’t understand it but so many casual restaurants do not want you eating in them anymore, and it’s not just the big chains.
It’s confusing to me. Does nobody else have business or a fun activity in another neighborhood and stop for a quick bite anymore? Or meet up with friends at the cheap restaurant down the street?
Might it be that the restaurants are responding to society not wanting to eat in them?
I recently moved back to a city I lived in before Covid, and the biggest shock is how most eateries survived the pandemic, but they have few actual patrons now, only a constant flow of square-backpack delivery drivers going in and out of them. I'm told it's because people in their 20s and early 30s now (i.e. the demographic here with a lot of disposable income) aren't much interested in going out to eat; it has become normal that the food comes to them.
The difference is moderation and the sheer amount of bots swarming everything. You can't put anything up a website with user input without paying for 15 different services to shield you from the hive. A website for people to talk to each other? You'll be arrested after it turns out adults use it to talk to minors. You want people to post images? Oh god I hope you're paying for state of the art content moderation LLMs, or risk traumatizing your most dedicated users who volunteer.
There's opportunity out there to build the tools that enable protection for smaller, more focused internet. Until then, its a dark forest.
I don’t intend to argue against your main point but for anyone who it might be helpful to, OpenAI has a free moderation API. A few years ago moderation was either something you did by hand or got volunteers to do.
That's a useful tool, but given how much OpenAI themselves is one of the organisations hurting the web, it's not something I'd want to rely on.
> A website for people to talk to each other? You'll be arrested after it turns out adults use it to talk to minors
Adults talking to minors of the forum wont make you arrested. But, blatantly facilitating child abuse again and again and again maybe with some low probability will.
The two are not the same. Not at all.
Eh, there are plenty of forums out there doing just fine, and the same with other sites like wikis, image boards, chat rooms, etc.
None of them have needed to sign up for any sort of AI image classification service or moderation setup. My own site has done fine for the last decade with just human moderators and a relaxed community.
The personal web. Public entities, like governments, companies and non-profits still put in the effort to run their own websites, of which there are millions.
It's hardly surprising most normal private individuals don't want to do this and prefer a Web 2.0 style hosted app. They offer better convenience, design, privacy, security, etc.
The layoffs will redistribute it. Builders will build.
Build it and they will come doesn’t work well for websites.
Nor for anything else.
I feel it's pronounced online. I'm a big gossip when it comes to new businesses coming into the neighborhood. I notice a tenant goes out and then I'm watching and asking people if they know anything about what's coming in. Everyone I know seems to be doing something similar. We may not visit if we're not interested in the goods/services, but we're aware of it if the need arises. But websites launch in the dark, practically invisible. If I search for it's exact good/service, no chance I'll find it (even pre-AI). It's pretty much a field of dreams with a sad ending, or a whole lot of work and money to buy traffic. Even some of these big Show HN's that spend a day or so on the front page, I bet most of them go back to ~0 real fast after that. There was certainly a time where that was like a winning lottery ticket. Maybe I'm off the mark, or cynical, I'd love to hear a story of a recent example where Show HN turned into a big MRR in a short timespan.
I forget the name but there once was a short lived search engine that heavily leaned on crawling new domains.
I only ran a dozen queries but it was memorable. I didn't see a website with "under construction" in it in the decade before nor the decade after.
It seems like an unfinished approach. (Perhaps add dns data?)
Someone should do a p2p search engine and crawl those 100 to 800 thousand new domains per day. Leave them in the index for a week or a month. Make the removed ones into a download, perhaps with some [traffic] data.
Then have the user set up some queries and give them a notification if a new result happens. With suggestions like `$city $restaurant`
Problem is, nobody cares enough to do that. To use that, I mean. These things were neat in the earlier Internet when it was a toy and we were seeking information. Now we have too much information.
The whole StumbleUpon idea was huge for a while too. Didn’t age well though, it was a novelty. I think this was around the time most people switched to a Feed style of information and that sucks your available attention
If you haven't tried it you should spin up a yacy just to see how much fun it is.
https://yacy.net
The once beautiful visualisation of nodes has turned into a glowing donut of activity. Lots of people care enough to use it. You don't have to use the network, it's already great if you only crawl the website you care about.
Ideal I think is to give it some bandwidth the first few days then tune it down to the minimum.
No one is building on the public web today though.
Everything is a private app, or walled-garden served over HTTP/s.
The web now is mostly an ecommerce platform with Google being an e-commerce search engine.
The web of late 1990’s early 2000’s is dead. Sadly,
HTTPS is a great thing and orthogonal to the issue
My point was more the protocol of the web, both http and https is now used to serve walled private gardens not public websites.
I probably could have worded it better.
Indeed HTTPS / TLS is a good thing.
Actually, it isn't. We should have many protocols in the browser.
With enough gate keeping things could get to a point where the peasants have to extend a document format to be able to publish applications.
When that happens we will have to prevent users from finding them until we find a better method of shutting them down.
One day we will control what they can talk about and our official narrative will be all that remains.</sarcasm>
My sector is currently on strike, I could find exactly one outdated news article where one of the big employers says it has no influence on operations. Like they would pay many millions per month for work they could easily do without.
We still have some resistance from heroes of the old days: https://best-electronics-ca.com/800.htm
Is writing happening anywhere? That's my big fear.
The last two surviving text based sites are reddit and x. Reddit is going out of their way to become an Instagram feed, and what remains is mostly bots and astroturfing. X, even ignoring the bots and nazis, is not a place for long format.
There's still pockets like IMDb reviews or this site, but it's an extreme niche. I even have people at work who communicate by audios and videos whenever they can rather than writing on slack.
Substack.
They seem to be having AI spam issues though. I came back after a half year away to post an article and the interactions seem like bots
There will eventually be small, highly moderated corners of what was once an open internet pop up. They probably already exist. Nothing of this internet serves people, the people serve it. That will change eventually.
I think this is true, probably even good, my worry is I wont be able to find them or be let in :(
AO3 just came back up, but its absence was heavily noticed and memed. Human writing for entertainment is still relevant.
I was about to mention this and noticed your post. Not my cup of tea, but my niece was holding her breath for it to come back up for the full two and a bit weeks it was down. I told her to prepare for the worst, that length downtime is usually fatal, but I think it was the day after I said that when it came back online and a great weight was lifted from my niece's shoulders. She's Gen Z and loves the written word. Little darlin'.
start a forum, my man! there are dozens us of us, still!
Telegram, in many ways, reminds me of forums of old. Feeds into the centralization of everything, but well...
If you sign up with a voip number you often get banned after a few hours/days
IMDB is owned by Amazon and is getting enshitified.
Imdb became crap within six months of Amazon buying it.
Substack
Matrix and IRC? ActivityPub (ie mastodon)? ATProto (ie bluesky)?
Substack? They’re trying hard to ban AI content and focus on long format.
All the good writing is on Substack.
Mostly in its own domains. My niche interests are covered in sites their readers pay for, so i built a search engine to find such. Quite many sites supported with Patreon, Substack, Ko-fi etc. Shameless plug: readerfunded.com The index is not very big, but it helps me find interesting content.
Google is responsible for adsense and they decide who gets search traffic and who doesn't. They are complicit in the enshittification of their own game.
They've made the internet broadly unuseable. I absolutely loathe visiting unfamiliar websites. Sometimes you can't even find the article amidst the spam, pop-ups and autoplay videos. Remember when we used to shame users for commenting before reading the article? Most of these sites today you wouldn't wish upon your childhood bully, of course nobody reads the articles anymore.
Meanwhile I've conditioned myself to not even recognize what it is that's being advertised at me, so it's all for naught.
It's a futile arms race between advertisers finding ever more annoying ways of getting my attention and me finding ever more devious ways of averting my eyes from whatever it is they want me to recognize.
I suspect Google drives just enough traffic to crappified websites like Pinterest, Yelp and IMDB to keep them on life support so it can avoid anti-trust lawsuits. That's the only reason they exist.
You can argue that since AI there are now more players overall.
well.. that really wasn't my experience. when you put "who is on snl tonight," you would get a top row of news stories related to that, probably in this case some (admittedly low effort) post saying "Catch Tom Holland on SNL this Saturday!"
that said, it's true that SEO did kill even that, because the headlines began to be "You won't believe who tonight's SNL headliner is" and they even keep the name out of the slug too.
# You won't believe who tonight's SNL headliner is.
## Someone is hosting SNL this week.
In this article we will tell you who is hosting SNL this week.
### What is SNL?
SNL, or Satruday Night Live, is a sketch comedy show that originated back in...
### A brief history of sketch comedy
...
### Past SNL hosts
...
### This week on SNL
<this section still does not say who is hosting this week>
...
...
...
...
---
This almost every webpage/blog post I've found through a search engine since around 2016 (I remember around that time every single time I'd search something about React, every article would start with many paragraphs describing what React is).
Anyway, ya. People say that nostalgia for the "old web" is just that (nostalgia). I disagree, but ya.
Interesting...
When I seached it, one of the top links was this very same Hacker News thread.
> and so it goes...
Is there anything after AI though? As long as we allow AI to benefit individual people, we seem headed to an ad-free custom GUI future.
As you say, it has already started with AI rewriting web results into a focused text response--a text interface.
Frontier AI models have started drawing little diagrams and creating little interactive apps that go along with whatever they're trying to demonstrate. They also extract images from--well, from somewhere, the web I assume?--and placing those images in their text responses.
I've been using AI to write custom widgets for the game Beyond All Reason (which is open source and allows custom widgets), and it's great even with this obscure API. It can turn ideas into beautiful custom graphics easily.
I don't want to see your ads, or even your webpage. I just want the information.
Maybe with AI everywhere we can finally do some sort of micropayments where the AI can distribute a fixed amount of money on my behalf. I am willing to directly pay a small amount for casual browsing.
> ad-free
snort, No, you will get ads. It will be all ads eventually. It always turns into ads.
My computer is already capable of running an AI locally that can strip ads. Why would that ability stop?
Everyone's already on thin clients, I don't see that reversing for the average person.
One of the most depressing parts of life today is how much the thin client modality has become pervasive!
IMO it's a downstream problem of computers shifting from "specialized tool that responds to the owner and gives them agency" and towards "something effectively owned by a bunch of competing companies to push whatever they decide into your face."
The latter can be "thinned" much more effectively.
I guess the issues is where do you get the data from?
If you ask "flight from x to y" and your search provider only shows you the company willing to pay, what do you strip from the results?
Then I guess this whole "free market" experiment is pretty much over, because aren't there several companies who offer flights? And those companies still want to attract customers right (I mean, if companies don't want customers, what are we even doing?), so why wouldn't other airlines offer information about their flights my AI can collect?
There may be a niche case where only one airline offers an unusual flight, so be it, but in general there are always going to be multiple companies offering flights, and each of those companies wants to participate in the market (that's kind of the point of our economic system) and so they will make information about their flights available.
The airlines will, but will the search engines?
I'm saying you can't "filter out ads" if all the content you get is already ads, and you can't replace a search index with a local LLM.
You're right, but again, if people want to find flight information from multiple companies, but no company is willing or able to offer customers what they want, then why do we even have companies?
If push comes to shove given local AI it might well turn out that we don't need search engines to the same extent that we used to. Or rather that much of the process can be run locally.
Of course that would burn through a lot of resources and generally be inconvenient in various ways so inevitably someone would launch a centralized version and the cycle repeats.
> because aren't there several companies who offer flights?
That depends very much on where you are going and how much time you have to wait for connecting flights.
> And those companies still want to attract customers
If they're the only game in town they don't have to attract them. Airlines would rather consolidate and collude than compete. https://apnews.com/travel-a6ad554f766540dba6536c7d41ab0d39
> so why wouldn't other airlines offer information about their flights my AI can collect?
Because they'll want you to use their app to get that information so that they can collect massive amounts of your personal data and customize what flights and prices they'll offer you depending on who they think you are and how much money they think you have.
The "free market" means that companies are free make you use proprietary AI that shows you ads to access information about their goods and services. It also means that companies are free to make deals to collude about this. Same as how ticketmaster has deals with all the venues that says they have to use them for ticket sales. They are free to screw you over.
The only way to strip all ads is to deliver no information. They're not going to all be as obvious as "If you're hungry, why not try the new Bacon BaconBacon from Wendy's?"
Imagine you ask "quick snacks for Superbowl Sunday" - but the dairy industry has bought ad space. You'll get suggestions for cheesy nachos and spinach dip for sure, and chicken wings and other non-dairy options might be slightly less likely to appear, or described in slightly less enthusiastic tones.
And if the AI folk have their way, it won't stay in the chatbot either. When your day-planning agent or personal shopper gets ads, you won't even see the text.
Sorry for offtopic, at his point I more surprised (or just didn't notice) why there no local llm projects with browser plugin to filter/categorize HN content with headless entering site to analysis content. for example : topic with links, papers and algorithms inside, DIY style (no vendor locking) tutorial and so on.
It'll also be worse than before, since:
1. "Ads" through LLMs are so much easier to personalize to the viewer and can be deployed at just the right moment to influence them in a way that might pass as organic.
2. Companies/politicians/wealthy people can simply pay for a model to be generically favorable to them, where positive words/events are mentioned and negative aspects are given dynamically-generated excuses or buried. (Or unfavorable to opponents, if the model-company thinks they can avoid lawsuits over it.)
3. Broad bias is easier to purchase, where a consortium or political party can pay to favor one entire class of thing over another, without developing individual ads for specific products or candidates.
In the end, the risk-profile is similar to hiring a human "personal assistant"... who takes secret bribes to manipulate you.
You can't download a Google or Meta but you can download a local model. This time is different
Yeah, just like everyone hosts their email server. Techies are hopelessly optimistic.
As an individual user, there is no reason on God's green earth to host your own email server. (No. Sorry. There's not.)
There are multiple very good reasons to host your own AI models. Those reasons are not going away with time, but rather multiplying.
It's not about "reasons to host".
It's about capability:
1. Do you have the right hardware? Most people don't and hardware has just exploded in price. One of the RAM makers and also AMD have said they will focus on data centers, which will only make things worse.
2. Do you have the know-how? It needs to be as easy as installing and using Chrome.
3. The end result needs to be directly as usable as ChatGPT.
Fairly local LLMs fail all 3 points.
Cloud-hosted remains more efficient in terms of electricity, RAM use for large GB models, etc.
However, Apple is already including LLM models automatically with even iOS that can be queried offline for historical facts, etc.
I'm half optimistic about this, but we're so far from it from a hardware point of view.
We need unified memory for almost everything for these types of setups to be usable and we also need to standardize LLMs, if everyone wants to run their own LLM locally, that will probably be untenable for decades. We probably need new DRAM tech (3D RAM?).
Hardware prices need to go down a lot, and we're probably 5+ years away from that. Not only do prices need to go back to where they were, they need to become even lower. So we're probably 10+ years away from usable good enough local LLMs for the masses.
It's going to be an ugly decade from a hardware economics point of view. The potential is endless, but the way there is hard.
They will outlaw the open models. Probably very soon with safety as the excuse as always.
All the worst outcomes involve taking this technology away from normal people.
My dude, I get where you're coming from, but we've got decades of disappointing evidence that things will go a different way, absent something new like a political movement with legislation.
The predominantly average user isn't going to even know what the difference is between local and remote models, they're going to use whatever promises "fast and instant" access to read and modify the data that was already captured by "cloud" providers.
And that's not even starting to get into issues like figuring out which local model is actually trustworthy, picking something suitable to hardware available, how to hook up all the read/write access, etc.
The next SEO might be model poisoning. Your LLM will happen to consider coca-cola quite refreshing.
Google was ok with SEO. Or they'd have penalised those awful websites.
I'm sure some people will try to game LLMs to get it to promote what they want, but it's only a matter of time until the AI companies offer manipulation and ad injection as a paid service.
> Is there anything after AI though? As long as we allow AI to benefit individual people, we seem headed to an ad-free custom GUI future.
Every time I do a google search to cross check something, the AI answer has been incomplete or incorrect.
I can only assume the answers I don't know are the same.
I don't want the AI answer, I want the actual information.
> When you typed who is on snl tonight? You used to get 10 pages of clickbait that made you scroll down endlessly for the answer.
Right. Why? That's a nightmare that Google built for us. They lined the pockets of SEO spammers, and because it was their business, they talked the talk but were never willing to truly deal with the problem.
But even in your SNL example, to show you the answer, Google AI needs to take it from someone who actually made an effort to put this information online. Network programming is an rare corner case where the person doesn't mind the bot slurping it up because the goal is to get people to watch SNL, not to visit a website. But there is so much non-commercial content on the internet that is really getting hosed right now, and that people probably simply won't be putting online anymore because they don't want to do free labor for OpenAI or Google.
As a practical example, there's a lot of hobby historian communities dedicated to specific places, periods, or types of collectibles, and the only reason they exist is to bring people together and share their finds and wisdom. If you're a historian of Kinder Surprise toys or old military uniforms or whatever, and Google is now taking your work without letting you connect with like-minded individuals, there's really little incentive left.
Well, I'd say if people want communities, they'll always find a way to build connections—that part hasn't changed anyway.
The real difference comes down to who's paying the bill for all the server stuff. "Clicks" bring in ad revenue, which is what kept the internet "free" all these years, but that model might be coming to an end.
Nothing really free in the beginning after all.
Those niche special interests will always have hangout spots. YouTubers putting out weekly videos, subreddits, discord servers, facebook groups, irc channels, conventions. As long as it’s something popular enough to have a certain critical mass —- maybe one thousand dedicated fans world wide —- there’s going to be places to hang out.
Who cares if it isn’t googleable. There’s other ways to build community and other avenues to be recognized for your work.
No ? They never had hangout spots, or found critical mass, until the internet came along and made finding people wildly easier.
AI search results are inserting themselves as intermediaries in that process of discovery. This is the opposite of what the net achieved, and its reducing the rate at which potential group members ever find community.
YouTube in particular is already getting hemorrhaged by ai content.
I don't think this will play out as you imagine there, because those platforms will drown out human made content with ai copycats until nothing but the already established players are left
Its not that hard to avoid. Whenever you see an AI screenshot, click dont recommend this channel again. After a few days they all clear out and you can go back to the traditional endless feed of nothing but low quality human made garbage that has nothing to do with your interests, just as God intended.
AIs are pretty good and helping you avoid AI content.
The search box on YouTube is worthless and searching from Google or Bing is almost as bad. But if you ask Claude for a non-AI video that explains X, it does a good job of finding it.
The real pro-tip though is to use YouTube with intention. Don't go there and start scrolling. Go there for something specific and leave as soon as the video is over.
Will it though? How many people are watching AI videos and subscribing to the channel?
> Who cares if it isn’t googleable.
It's funny. Until a couple years ago I really hated walled gardens like facebook groups, discord and other app-only communities. But now I find myself defending them, as a last frontier against automation, where you can still hangout with real humans online...
How do you know those "humanos" arent bots?
Depending on the group, because you know some personally who know others personally ... a chain of trust. In other cases because you at least know some online long enough to be quite sure they are human. Also most bots are still spottable, just not always on first glance.
Who cares if it isn’t googleable
The difference is that Google aggregates content from everywhere, so you don't need a pre-existing relationship with any of those places to find content. If you want to find a YouTube video without Google you have to go to YouTube to search for it. That's great for YouTube, but it makes it virtually impossible for any other video platform to get a foothold.
The beauty of Google is (was?) that it was a way to get traffic if you weren't already the incumbent. What you're suggesting is that we should give up on having new platforms and just stick with the existing ones, even if they're terrible now or they decline in the future. That's not fun.
Google wasn’t that important for the web. Plenty of the things I’ve learned about for the first time in the last decade haven’t been driven by Google, because the top Google results are mostly SEO nonsense.
Google was great for finding products and services, but poor at finding the type of communities we are describing.
This feels very much like a personal anecdote, I can't see how you came up with "Google wasn't that important for the web" otherwise
The window where Google was an amazing way to find anything on the web was, at this point, quite small, on the other hand the things like AMP, killing off Google Reader, SEO and promoted listings were all detrimental for the web. So important, yes, but a net positive, I'm not sure anymore.
Google was great for finding communities, too. You're just younger than me.
In my personal experience Google’s ability to find communities started declining about 2010 or so, when all of the web forums started migrating to Facebook groups and Subreddits.
If they decline in the future they get replaced, as Google replaced Yahoo before it.
Indeed. I average 1k page views/day, I would guess 300-500 visitors/day. I'm fine with that. For one thing, I can personally respond to every comment usually in real time.
Google directly and indirectly benefits from SEO spam. They are a solution to a messy web, the messier the web, the more necessary they are. Without web spam, Google would have just been another search engine next to Altavista.
Google is also in the unique position of selling both search ads and display ads. And they know what a visitors ad exposure is for links in the search results. They make money on both sides of the click.
SEO is not a general messy web though, it's a specific attack on Google and Google search users that makes Google search less useful. I think it's quite a stretch to say Google benefits from it.
There’s a massive conflict of interest when Google owns both Search and AdWords. Google directly made a profit from SEO.
Search ads make Google way more money than display ads on SEO websites. Damaging search would have to have a huge payoff to be worth it.
(I do work for Alphabet so make of that what you will, maybe the Upton Sinclair quote does apply.)
Without SEO spam it would be much easier to create a google competitor so they benefit from it while actively fighting it at the same time.
> If you're a historian of Kinder Surprise toys or old military uniforms or whatever, and Google is now taking your work without letting you connect with like-minded individuals, there's really little incentive left.
Won't the people who are interested in connecting with like minded individuals click through anyway? The 40% who aren't clicking now are the people who were just coming to get the info and bounce, they weren't the ones connecting with like minded individuals.
If the sites go under because of this, there won’t be a site to click to in the first place.
Sites that were too big to just be run as a couple hours a week hobby project level, for years, relied on all traffic, including people who only read a page or two, to generate money via ad revenue. And the site being big, and existing in search results meant it was there for that core audience of people who might build community there.
All this is undone if hosting a site is no longer rewarding because you get no actual visitors and no community forms, while your content is just repackaged by commercial AI.
And I say this as a definite net promoter of AI. I just am honest that from a content publisher’s point of view, yup, they are getting screwed hard here - there’s no upside for them.
Yahoo started off as a directory. Maybe it is time to bring back curated directories again?
As an HN oldtimer, I can recognize your post as a typical reply made 10 years ago already, and as we can see, that never went anywhere. The web continued to decline.
The biggest problem is the general shift by which most of the world uses smartphones to access the internet, not computers. This medium seems to discourage traditional web content at a societal scale. A lot of younger people barely ever open a browser, their internet use is exclusively through apps. That causes traditional web content to gray out and become even less appealing to audiences.
>... smartphones to access the internet, not computers.
I was surprised when I learned that my obscure blog (no app; 1k page views/day average) gets 70% of its traffic from smartphones, congruent with your observation.
Since I solely use my laptop to produce my blog I was taken aback and immediately checked to see how it looks on my phone — something I'd never done.
I was relieved to see it displays perfectly there: in fact, it looks much better.
Most blogging platforms have had decent mobile theming for long years now. All the same, the small format of a smartphone screen seems to discourage, on a societal level, longform text reading, hence the expansion of microblogging, then image content in the Instagram age, then short-video content.
>... the small format of a smartphone screen seems to discourage, on a societal level, longform text reading...
FWIW just for lulz the other day I tried the iPhone Kindle app to see what it was like to read books there. I used white type on black background: excellent! Easier, sharper, all around better than my latest gen Kindle Paperwhite, which is what I use to read in bed at night.
I won't use screens* (read phones/tablets/laptops) in my bedroom so I'll stick to the Paperwhite there.
Yes, the Paperwhite has a screen, but e-ink is a whole different genre than the others.
This perhaps puts us in a situation where we return to the days of web 0.1... People put things online because they want them to be accessible online and for no other reason. Not for ad revenue, not for eyeball clout... Just because it's more convenient to have it online than stuck offline.
> People put things online because they want them to be accessible online
And the costs of putting something on line were commensurate with the number of people who needed to be paid to accomplish that goal. When Al Gore made the first website in the early 90s, it probably cost a few pennies of electricity and the effort of some grad students who worked for a few more pennies. Today, putting a website online involves paying a lot more companies who have massive bottom lines to support.
FWIW Blogger is still free.
It really doesn't. No feature that Al Gore relied upon has been removed (with the possible exception of the blink tag).
I maintain a blog from a static served page. It's a fancy text file. Granted, it took me years of study to know what to put in the text file and I spend like a hundred bucks a year on hosting. But I spend more than that on a Netflix subscription.
You're thinking about putting a website online with a massive amount of functionality on top of it, or SEOing it, or monitoring it for traffic or ad integrating it or... All of that is web 1.0 and Web 2.0 thinking.
Web 0.1 was hand uploading bespoke HTML and possibly using a static site generator. Documents served by academics and government functionaries to other academics and government functionaries because the alternative ways to share those documents were more expensive.
What I meant is that, today, there are a lot more intermediaries who want to get paid, in between your html and your end user.
Like who? You need a domain, and a place to host it. That seems like the same set of things you've always needed?
? a static website is very easy to set-up in 2026, you can even ask AI how.
The problem is most such sites were so horribly overloaded with obnoxious advertising that they were unusable without an ad blocker. They had already chosen to make their sites unbearable long before AI came along, so it's hard to mourn them now.
You are not born with interests, you develop them over time. Often by discovering that there exists a cool community you can be a part of.
And that's before we get to the other problem: what's under the AI summaries, and what's often "cited" by the summaries, is LLM slop too, because it's now practically free to create on any topic. For example, I searched for info on a vintage calculator just yesterday, and ended up on this page:
https://www.calculator.org/calculators/Calfax_Calfax.html
It's just data scraped from real hobby websites to boost their page rank. And everything is like that now. No subject is too insignificant for LLM spam because the cost is negligible.
I’m sorry but there is no one needed to post who is on SNL tonight, nor when certain sporting events are occurring, nor when the next episode of $SHOW comes on.
These are far more common than niche hobby collectible communities, and if those communities love bringing people together those people can still hang out together!
People who are looking for others with a similar passion—which, if it’s a niche interest, sounds quite nice—still have access to search results or the sources shown on the AI summary.
But for the vast majority of things, we don’t need 29 sites competing for SEO on “Which channel is the US Open on?”
I think there is probably some long-tail data, but I'm skeptical a large volume of questions actually ended up going to hobbyist historian communities in the first place. Unless I'm misremembering the before times, like 99% of my history-related queries on Google led me to Wikipedia. I might as well have been doing a site:wikipedia.org search for how reliably that happened. Now, the AI Overview just summarizes the same Wikipedia article. Not a huge change in that particular niche imo, unless you're tracking Wikipedia pageviews. Which even as a longtime Wikipedia editor I don't care about (I don't edit it to rack up pageviews).
Well that will at least undermine the point of Wikipedia, which is that readers can edit it and improve it. If people are reading it only through AI summaries, Wikipedia will eventually lose that critical mass of editor community that made it possible in the first place
> Well that will at least undermine the point of Wikipedia, which is that readers can edit it and improve it. If people are reading it only through AI summaries, Wikipedia will eventually lose that critical mass of editor community that made it possible in the first place
IMHO, Wikipedia has similar culture problems to Stack Overflow, just less extreme, and its editor decline trend started before AI and AI overviews.
Due to those culture problems, eventually Wikipedia will be completely dominated by people dedicated to pushing ideological agendas and others who just like throwing the book at newbies.
Most of those AI summaries are pulling right from Wikipedia and their answer accuracy is calibrated against it. If anything, the problem is going to be that Wikipedia answers become an unstopable circular source of truth.
wikipedia does not allow AI as a source
different thing is that it allows joirnalism... that can quote wrong from wikipedia
this topic is in my opinion much smaller than AI slop learning om AI slop
for the impact of wikipedia on journalism (mankind?) the circular reporting cases are relatively minor
https://en.wikipedia.org/wiki/Circular_reporting
https://en.wikipedia.org/wiki/Wikipedia:List_of_citogenesis_...
I would complain more about
1) death of forums
2) reddit being plagued by bots, artificial upvotes
3) reddit mod problem (mods can create a community, but also mods cannot be removed. removal would be a problem due to brigading, but there is no mechanism to remove bad mods)
4) HN becoming reddit / eternal September
> wikipedia does not allow AI as a source
Maybe not directly, but what about sources that were AI generated? How would that even be vetted?
Wikipedia has never been a community of occasional contributors. It's controlled with an iron fist by a small group of insiders that treat the wiki pages like their personal fiefdom and often have no personal interest (or expertise) at all in the topics that they write about.
This may be the impression for people that are not familiar with how Wikipedia works (and that's a problem in itself), but in my experience (20+ years of contributions) this is not true at all. The community is very open to discuss different ideas and it always surprises me how opinions and consensus can quickly change with new information from reliable sources. As for the problem of seeming hostile to newcomers, this is due to the rules that govern the editing process in the community, the result of years of debates. This is in my opinion inevitable in order to run a successful community, but it also means that it requires some effort to join it
A good amount of Wikipedia pages are controlled by editors with autism (sometimes disabling to the point of inability to hold a job, hence so much time to edit Wikipedia), and indeed they have a personal interest in the subject that amounts to outright obsession--and that's the problem, that's why they can sometimes perceive other people's edits as a threat.
That would be a case of [citation needed]
In this case it's the editors I interacted with (myself included) closely through long years of editing Wikipedia. Now, I moved on to other things before Covid, but with the difficulty of crowdsourced projects to ensure turnover of a healthy general public of contributors in the smartphone age, I can't imagine things have improved from this in any way.
>> in the topics that they write about.
Any most of them haven't done any real writing in years. They stand atop their domains angrily fighting off anyone who dares to update something without kissing the ring. They don't do research and only contribute through deletions and reversals of other people's writing.
I have a few relatives infamous enough to have Wikipedia entries. Every time I have tried to update them, a mod has pounced. Where my grandfather died ... nope. That cannot be changed without me submitting heaps of personal information to some rando mod. And good luck going anywhere near article about legal topics. Lawyer? Published expert in your narrow field? Not good enough. It's armchair lawyers only on wikipedia.
Wikipedia has problems with territorial editors, but you aren't going to get much sympathy when being rebuffed for editing pages based on personal acquaintance with the subject. That a higher standard of citation is required than "Trust me, I know, I'm a relative" has been a basic feature of the project since its founding in the early 2000s.
Which is a totally reasonable thing, but the amount of required evidence is arbitrary and dictated by angry mods. In this particular case, my grandfather had stepped out of public life decades earlier. So to prove that the "John Smith" who died in city X was the same "John Smith" who was famous in city Y, the mod wanted personal ID from me and an entire life history linking the famous "John Smith" to the dead one. Just... no. I am not going to draft a multi-page biography and hand over reams of family info to a rando username. My grandfather's one-page Wikipedia entry will just have to remain wrong, forever.
The other small change I wanted to make was adding a note stating that several of his listed children and grandchildren (ie me) had continued the family tradition of military service. Again, the mod wanted way too much PII. That small bit of color will also never appear in wikipedia. It is no wonder that famous/political families hire PR firms to negotiate and update their Wikipedia listings.
How is that any worse than citing some random newspaper? At some point _someone_ has to cite a primary source that isn't available online. Wikipedia is just kicking the can down the road.
An encyclopedia is a tertiary source. A random newspaper is a secondary source. It has always been the case that the former uses the latter as citations.
This reminds me of my experience some years ago with Reddit. I volunteered to host a real time Q&A about anesthesia (I'm a retired neurosurgical anesthesiologist). After about an hour of photographing and uploading every license/certificate of board qualification/evidence of continuing education requirements fulfilled/driver's license/utility bills/articles I've published/Google Scholar page/passport, they said my proof of identity/qualification was insufficient so my offer was denied.
“Non-commercial content”. I’m not sure this is a thing anymore.
It's very true. I am hesitant to post things I think are novel now because I know it will just go into training data, unattributed. I even comment on HN a lot less. It's wholesale ripping off of people's IP - the chilling effects of this will not be felt for a while.
Whether it goes into training data is irrelevant. The AI agent (whether embedded into google's search or otherwise) will do the search on behalf of the person, find the answer in your page and return the answer, rather than your page. It's the ability of AI to search and summarize that has changed things. Not whether the LLM itself has absorbed your particular information. Once LLM's were good enough to make reasonable tools calls to search the web it was game over. It's just like how calling a library to ask a factual question of a librarian, rather than having to locate the information in a primary (or secondary) source yourself disintermediated the encylopedia publishers. It's just that calling librarians was never particular efficient or practical.
It's completely wrecked the internet for me.
Between SEO spam as the precursor then AI generated SEO spam and then just AI.
I used to enjoy stumbling on random things and learning/seeing into a world I didn't know existed (usually related to tech/science/history) but the signal/noise has changed radically and not in favour of the signal.
I think the problem is that we depend on Google or other mega-corporations for information-discovery on the world wide web. It's not just a search engine - it is now top-down control by these private entities.
We need a different version of the www. Google controls the www though via the adChromium codebase.
The only actual good parts of the internet left are extremely walled off, gate kept places with a handful of people who know each other. Everything else is a completely lost cause, including this site.
Could you give some concrete - or abstract example what these places would be (like)?
Not the parent poster, but in my experience it’s things like invite only forums, discords, irc servers, blog rings, etc.
Jeez, I didn't even think of providing free training data. I've been more worried about data harvesting and attributions to some sort of marketing profile (or worse).
>It's wholesale ripping off of people's IP - the chilling effects of this will not be felt for a while.
The wholesale ripping off started in the late 90's and we are already deep into the "feeling it". DRM, standalone apps for everything, consolidated corporate internet with paywalls all over.
I'm sorry, I just cannot take anyone on the internet who bemoans IP laws seriously. We all ad block everything and pirate stuff as needed, this is the moral high ground of the internet for 25+ years. Hell LLMs are the worst pirates anyway because they won't even let you just cleanly download the IP you want to steal.
Copyright is only useful for rent-seeking if you have a popular enough IP and lawyers on speed dial to collect that rent or for censoring the availability of IP. DNS blocks are a type of copyright censorship.
Copyright has little or no value for small blokes and bloketts like us. I think that part of the problem with copyright is that we are blending rent-seeking with acknowledgement of one's contribution in the creation of IP.
I dare say, most of the value of “copyright” for us small players is that it is part of the discovery process through which people with jobs and money find us. My writings and code have led customers to me (in a good way).
I don't mind what the LLM hovers up of my so-called creative efforts as long as people can find me. And that perspective is, I think, the core complaint people have about Google's AI searches. At least, that's how I read the article.
FWIW, having known the people responsible at Google for dealing with SEO spam...
It was never a question of resources and it was never a conflicting goals issue. From Google's standpoint, SEO spam is an existential threat because if somebody solves filtering for it, everyone will switch to that competitor.
It is simply a hard problem to solve at scale. Successful spammers live on the razor line between the signals Google has for useful content and wasting people's time.
The element Google can't control for is human nature. SEO spam sites are built to keep organic users reading long enough to tip the quality signal in the site's favor. Savvy users like those in this conversation know what they look like, but remember that most of the internet is not composed of sanity users. So the signal Google mostly gets from those pages is that they are holding user attention, and holding user attention, and holding use your attention because the site is annoying to retrieve information from is difficult to distinguish from holding user attention because the site is good.
First result: https://imgur.com/a/1KQI7Pj
> that people probably simply won't be putting online anymore because they don't want to do free labor for OpenAI or Google
I have bad news about their private anti-AI Discord communities.
It’s a push to do cheap labor for Google and make YouTube videos instead of text
I still keep my own website devoted to most of the interests I have, and sure, those opinions and information are destined for aggregation, but at least they are still worth something and reaching an audience, even if they are not the way I planned.
Once you've created something, whether it's an album or a blogpost, they are like your children - or as Kahlil Gibran wrote:
When people's interest is piqued, they may investigate more deeply, and that may lead them to other people in the same way. Discord servers are doing pretty well.
Personally, I'm a follower of Bowie's definition of success, in that success is about how much you influence other people, because it's purer and happens even if they hate you.
Monetisation is just one possible consequence of creation. We've somehow mistaken it for the purpose. Because capitalism.
Google could use the LLM to find the best hit in the SRPs and put those before the synopsis. It would take more context but if it has an idea of what the answer is it could review the text content of the top hits and check them against what it think the user is looking for.
Of course more often than not the reference links within LLM responses don't really relate to what they're saying - LLMs act like highschoolers, "I need X references to back up my essay, so I'll find some things that sound relevant and put them in as footnotes".
At first search results were great, then enshittification happened. It’ll happen to AI too.
AI is an app, but it's also open-source software that runs locally. Emacs was never enshitified--why not?
Almost nobody is using local AI. Almost nobody is using a computing device they actually control. Most people don't even have a desktop or laptop, only cell phones which further limits their options. Actual free local LLMs have the potential to be useful in a lot of ways, but it's only useful when they are powerful, with hardware requirements people can easily manage, on hardware/operating system they control. If you can't afford ram and your computer just logs you into a cloud account where your OS spies on everything you do and pushes ads your AI app isn't going to save you.
> You used to get 10 pages of clickbait that made you scroll down endlessly for the answer.
And before that, you would get the an exact, human written page for what you searched for, which is why Google was able to make money in the first place.
This is a Google search problem, and they could have defeated SEO sites if it weren't more profitable for them. SEOs didn't force Google to put 8 ads between the search bar and the first non-ad result, Google did.
Whatever Google is now, nobody would actually choose it if there were alternatives that competed on product, but there aren't.
We all live in duopoly land where every extremely high margin software and hardware role is taken up by 2 of 7 companies.
> Whatever Google is now, nobody would actually choose it if there were alternatives that competed on product, but there aren't.
There are dozens but they aren't good either.
It turns out that running a good search engine is very challenging and insanely expensive.
Companies make good products as loss leaders then monetize them later. People complain that the product was "enshittified" but what they get wrong is the original product never existed in any real sense. It was never a durable maintainable system. Most things require money to run well and if they can't make money they won't keep running well.
I get what you mean, and free services can't run on good will. But an alternative read is that once money becomes the goal, things usually get worse.
There are three paths:
1. Money
2. Goodwill
3. Threat of violence (usually via government / taxes)
You can have some mix of those three (a mix of 80% money 20% goodwill seems pretty solid in my experience), but I don't think there is some secret fourth option.
Censorship & propaganda
I don’t mind paying money for a good not enshittified thing. Today I spent $20 on dumb AI video tokens. It was fun.
Google would never have existed without money being the goal.
Google was born out of CIA and NSA research grants - US intelligence services saw the potential of the internet as a platform for mass surveillance, data collection and psychological warfare back in the early 1990s.
As with the CIA funding the modern art movement and various counterculture groups, and the LSD that drove hippie culture being a product of MKULTRA, the money would have appeared from somewhere.
We used to call that a bait and switch scam, yknow. You post as though it’s inevitable, Google was profitable before 8 ads per result. They didn’t have to cannibalize themselves, they chose to
The fact that google ads went from obvious and clear, to nearly indistinguishable from search results, and took up more screen space over time, has nothing to do with building a good product.
It has everything to do with quarterly number go up. The market is meant to align incentives, and the incentives here is to transfer as much benefit as is possible to shareholders.
Google (Tech in general) makes more money on a smaller cost base than any industry that has ever existed.
> quarterly number go up
People say this and act like it has something to do with greed.
No. This is just basic economics. This is what it means to run a business with other people's money. If your growth doesn't compete with other places they could invest their money, why shouldn't they take it back?
> Google (Tech in general) makes more money on a smaller cost base than any industry that has ever existed.
This is quite simply not true. Tech does not have the largest profit margins of all time, and has exceptionally high capital requirements to operate.
What tech does have is really nice scaling properties, which Google benefits from as the "winner" in search. But beware survivorship bias.
Yes, that is why I said incentives.
Economics is the ordering of objective things to achieve subjective goals.
We like markets because they achieve several incentive alignments that result in better outcomes for humans - compared to alternative systems.
> This is quite simply not true. Tech does not have the largest profit margins of all time, and has exceptionally high capital requirements to operate
No, its very true and the basics of all industry valuation - here’s a ROIC comparison for industries:
https://pages.stern.nyu.edu/adamodar/New_Home_Page/datafile/...
I think tobacco is an industry that does better than tech.
Marginal costs of sales for tech are cheap, code scales like nothing that has ever existed before. I’ve worked in “legacy”, more capital intensive industries, where most of your revenue is consumed by your COGS. Tech has it absurdly absurdly good.
I am not sure if survivorship bias matters here; I can sort of see the value of the argument, but I am not entirely convinced.
It doesn’t change the operational logic of tech. Firms fail in all industries, so we are always looking at survivors when comparing. Competing is the basis of the market anyway, so the firms failing would be come under the cost of doing business.
Nothing basic about ad platform economics, they prevent new entrants, controlling demand and supply at the same time.
Did they prevent Facebook and Instagram from beginning huge ad platforms?
> People say this and act like it has something to do with greed.
> No. This is just basic economics.
A lot of what people call "basic economics" is just excuses for greed.
There's Kagi, and it proves a search engine can get rid of slop and ads perfectly well, if it's in their interest.
Kagi.
I get reliable search results the vast majority of times, and if there’s a site that keeps popping up that I don’t like (AI-generated slop or paywalled), I block it and my results improve.
You can block sites, because Kagi’s product is search, not ads. Yeah, you have to pay for it, but you know that? I’m fine with that. I’m fine with paying for a product to do a thing well, and not have it tainted by the alternate revenue stream the company actually uses to make money from.
I've bounced off Kagi a couple times, but they've progressively improved as Google has worsened, and now I think I'm here to stay. They're dramatically better at skipping past literal whole pages of results of slop copypasta sites now. Google is rather clearly asleep at the wheel.
Can you share a search on google where your first 1 or 2 results (excluding ads, because adblocker) isn’t a relevant/good result, but Kagi is?
I think this kind of exemplifies the problem. Yes Kagi is a superior product. Why haven’t they taken off? It seems reasonable to conclude that competing on product isn’t enough.
Yeah google is deeply entrenched and people have come to expect things for free.
I use Kagi, and I really hope I am part of a large enough cohort paying for it to be a good, viable business. But realistically it is incredibly difficult to imagine it taking any major market share given the above.
What's "taken off"? Not everything needs to be Google-scale to succeed - the world would be better with more, smaller companies.
Internet search isn’t one of those. Breadth, depth, and quality of search results (ie what you want is the first result) matter. That’s only reasonably attainable at scale. Not to mention it’d be a PITA to optimize for a dozen competing crawl standards.
You don't need a dozen crawlers, just a dozen good search engines that have access to an existing index.
Kagi is stuck at 75000 subscribers, when they should have at least a million. Google scale is unrealistic, but they haven't taken off.
No one outside of tech has heard of them.
Perhaps they should run some Google ads and work on their SEO.
Tell them, don't tell me! Yes, they are not doing much to reach low hanging fruit outside of tech industry people, even though a premium search engine has millions of potential customers.
I quit my Kagi subscription precisely because it's unable to keep up with the AI slop infesting every search. Even Kagi's results are mostly AI generated articles these days.
> And before that, you would get the an exact, human written page for what you searched for, which is why Google was able to make money in the first place.
Google wasn’t really good for questions like “who is hosting SNL?” until pretty late in the game. The Huffington Post made a big deal in the early 2010s about posting articles like “What time is the Super Bowl?” The SEO wars started pretty much immediately after that.
But before that you probably would have had to go to a TV listings site or go to the NBC page for SNL, which may or may not have had the answer, possibly embedded in a Flash interactive.
Prior to that you'd usually type "SNL" and check the Wikipedia overview. Or "SNL Wikipedia" if that didn't appear.
I remember that era. Honestly it was a bit more work but it wasn’t unpleasant work finding information. We 2000s-era netizens just learned the names of sites like IMDB, wunderground.com, WebMD, Mapquest, ZDNET… different sites for different purposes.
(Just as we had multiple big retailers, in different narrow categories, like CDNOW.com, before Amazon.) All the consolidation into Google for every information need is unhealthy at best.
It was always such a treat to find a new website that has information you were interested in.
I still remember how much to sleep I lost when I discovered TVTropes
This is why DDG is nice, for the !bangs
This is such a nonsense "conspiracy" style take. Google doesn't like SEO. If Google search sucks and Bing or duckduckgo or whatever else rules then everyone would switch over a few weeks. Google has whole teams of people trying to stop SEO from working because if they stopped giving better answers than their competitors then they'd be out of business
> If Google search sucks and Bing or duckduckgo or whatever else rules then everyone would switch over a few weeks
I switched, and every tech-savvy person that I know switched about 5 years to a decade ago. Google stays on business, just like Facebook, because of the huge advantage of being a household name.
Google is... google.
This is exaxctly how they built out their dominance in search. Good, high quality answers, no ads. Get the users in the door. Become synonymous with search.
Google absolutely bled cash for their first three years, because they were investing in acquisition. Here we are again.
Then, little by little, you monetise, until the product is once again awful.
They repeated this process with YouTube, for a larger example of "burn cash on user acquisition until vast profits".
We'll see the same with their AI offering. It'll take 10-15 years to the "hot trash" stage, it'll be WILDLY profitable (think of the targetting opportunities!), and then they'll have to pivot to the new hotness, whatever that is.
Whatever Google is now, nobody would actually choose it if there were alternatives that competed on product, but there aren't.
Yes, there are. I use ChatGTP Plus to search, with citations. It's not free, but my nerves aren't either.
It's not just death of discovery. This is theft. They scraped your website, built a knowledge base out of it with the promise of sending you people to you. But only now, they use it to answer the questions themselves and cut you off. Classic middleman play.
Yes you are right. However, over the years google search has exacerbated the shady websites that just exist to make a buck by SEO gaming and they even get their content by scraping other sites. Over time they became more popular than legit sites. So chatbots kind of put an end to those kind of sites.
In other words, the search traffic was already amplified by crap and this pulled it back. I am not saying there is no theft because there is massive theft.
No, you can't steal facts. Nobody owns the truth, no matter how hard you worked collecting them on your site.
Plus if anything AIOs link out more to source sites. Google used to give data answers to a lot of questions, which now are being answered with AIOs that cite websites instead. One egregious example I get a lot is sunset times - there's 0 need to cite a website when telling me when sunset will be, a simple python function can do that programmatically.
Intellectual property is a thing. Not everyone is running a Wikipedia. It's just a reductionist take since everything can be collectively labelled as facts.
"who is on SNL" is very clearly an uncopyrightable fact. _Most_ search queries are for uncopyrightable facts, or derivatives thereof (e.g. comparisons, recommendations).
> Intellectual property is a thing.
You can't copyright facts.
If you write a cookbook, I'm perfectly free to take all your recipies (facts) and publish them in a cookbook of my own. I'm not free to take the artwork and prose (intellectual property) from your cookbook without your permission.
https://en.wikipedia.org/wiki/Trap_street
https://en.wikipedia.org/wiki/Trap_street#In_the_United_Stat...
In Nester's Map & Guide Corp. v. Hagstrom Map Co. (1992), the United States District Court for the Eastern District of New York held that trap streets are not copyrightable under the federal law of the United States.
What happens on a football game is fact. How photons hit the field and the players and reflected into the camera is a fact. Now go set up a public real time football game stream and let us know how well "nobody owns the truth" defense works when you get sued.
Video and audio recordings/transmissions are treated differently under the law. You can't compare them easily to written works. Nevertheless, if you watch a TV stream of a game, and then use that to announce facts in real time, that's not an infringement. This is basically the same as what Google is doing with AI overviews; reading sources and extracting facts from them.
> the same as what Google is doing with AI overviews; reading sources and extracting facts from them
And was sued for it by news media in a number of countries like Australia.
"No one owns the truth" is exactly the kind of black and white thinking that leads to bad places.
> No, you can't steal facts. Nobody owns the truth, no matter how hard you worked collecting them on your site.
This is what I will tell to myself from now on when downloading movies and series instead of paying for them.
Facts != Intellectual property
They still own the rights to the work they made. What was stolen was a work, containing facts.
Ultimately it doesn't matter, because the internet will die, and Google will have no one to steal facts from anymore.
I provide a service where you call a phone number, I pick up, and you ask me who's on SNL tonight. I google it, visit the SNL webpage that lists their nightly cast, and I tell you the answer. Then I write the answer down on a post it in case anyone else calls and asks the same question. When I answer that question from then on, is that theft?
Not unless you think an encyclopedia is theft because it didn't come up with all its facts on its own
It's enclosure, not theft.
> the death of discovery is worth mourning
What you are describing is not the death of discovery. It is still required to discover a lot of things that are also more valuable to know.
For example you can't really understand how to implement _start on linux x86_64 by just reading some summary.
> You used to get 10 pages of clickbait that made you scroll down endlessly for the answer.
And this is why, kids, the statement:
"This technology is the worse it will ever be"
is false.
Whether it's search or cars with huge screens or LLMs. Technology can obviously regress, even if the on paper specs progress. It's all about incentives and who the actual main customer is.
> That said, the death of discovery is worth mourning.
Am I the only one discovering way more niche sites and cool research publications now, by clicking through the sources the LLM gives back to me?
AI didn't kill the SEO, AI is doing that SEO or something worse.
That thing exactly turned me off google search years before AI. Bing cloaked by duck duck go and similar privacy driven search engines seemed to be much less cluttered.
Now with AI (I use chatgpt and perplexity) as direct replacement for search engines I find Google as a company even more obsolete than in the past.
So is YT, which is cluttered with ads more than linear TV in American and I thought nothing could be worse… and recommendation engine focused on misinformation more than ever. All interesting content from YT once randomly found is ripped with yt-dlp and stored on NAS to be turned into local streaming service via plex or emby or similar service is my way to access media.
Google created habits of effective discovery of valuable content in the internet and killed it by its own greed.
I never understood the unilateral pre-AI attribution to AI for killing the web when Google search began ignoring search terms and providing irrelevant but thematically similar results years prior to Google force-feeding AI to all users.
The removal of `&&`, `||`, `(`, `)`, `"` having actual schematic meaning on google.com was really the canary in the coal mine. The system ceased to be a tool for information discovery and became one for ensuring the top results were ad driven.
And where will AI get its answers when the only contents of the internet are from low quality AI spam?
And when you clicked those links, you'd find maybe one wrong answer, and 9 good-enough or maybe even a great answer.
Now, you ask AI if Sonic Advance 2 allows you to get a D rank, it lies to your face that yes. The game doesn't even have letter ranks. Too bad you're streaming on a game show while searching that, letting 8 thousand people see the failure of the modern google in action.
> SEO killed the web
Google chose to prioritize results which don't provide the answer, making you scroll aimlessly and paying them with ad revenue.
Google chose to keep the websites with the answers for themselves, and only provide them through AI. There's a difference.
Google killed the web, period.
>You used to
You still do. And on the topics where the google does shove their AI in, the summary all too often is factually incorrect.
I so much hate google results with these clickbaits, and content farms. That is why I became obsessed with managing search by myself. It is not doable on big scale, but there might be enhancements.
So I have a project with domains, YouTube channels, Github repositories, reddit channels [0]. All have 'ranking', so most important appear first for me.
This project however is already big. 2M of entries. So I have a trimmed version of it, only things I already consider worth maintaining [1].
I also maintain RSS feeds for what is worth [2], and I have other repositories.
I don't like Google. They intentionally broken the search. You cannot say it differently once you understand how broad, rich, and interesting the internet actually is.
Although list of domains does not answer broad questions like "who is on snl tonight", so I still use other search methods.
[0] - https://github.com/rumca-js/Internet-Places-Database
[1] - https://github.com/rumca-js/awesome-database-top
[2] - https://github.com/rumca-js/awesome-database-feeds
Well, if you're really interested in the answer, Google AI also gives you the sources, so with any luck by clicking through you'll find a site that's a good resource for the topic, and that site will get some traffic. I do that regularly because I don't really trust Google to give me the correct answer, but I don't know how many others do it...
I hope AI also kills SaaS next to SEO. At least smaller businesses move away from it, larger enterprises will stay hostage longer or it will make more sense in their use case.
But yes, a lot of pipelines are broken right now. If you are a junior dev today, you have a hard time to get into the field. Quite ironic since just a year ago you were too old if you were past the age of 25 for any dev work...
They will probably come up with ‘SEO for AI’ and start injecting ads into their AI answers. That said, commercial web might be dead, however there are still enough people who host their own blog, have strict no-ai rules.
On the other hand decades ago web search was horrible - in most cases what you got was result of full text searches ordered by the rank. You had to go through many websites, read so much information just to extract essential knowledge.
Hyperbolic. In the past, such a query usually landed me on the right info within three or so links, until Google themselves started inserting pages of ads, "affiliated content", spam, info cards, etc. and encouraging tactics which favored tabloids and other high-churn websites. Google killed search.
Soon the Internet won't even have the answer.
You know, just ranking real sites with real information would have fixed that.
> SEO killed the web, AI killed SEO, and
Hopefully people go back to real social interactions and that kills AI on the web. People just go back to directly contacting businesses rather than exploring their what is now a completely AI generated website. Almost all the models used in the newspaper are now AI.
Why weren't you going to the SNL site? Anything else is just rumour.
> …the death of discovery is worth mourning.
SEO sites killed discovery a long time ago. And, based on what Google has been serving up for a decade [1], they did seemingly nothing about it.
Oh well.
([1] nothing worth discovering)
Yeah, IMO discovery was back when we had web directories, web rings, and stumbleupon. Search IMO killed discovery as you lost the ability to browse categories of sites.
For a good period, Google actually actively tried to combat linkbait, poorly built websites, duplicate content and other SEO shenanigans to try to reduce the art of SEO to "build a good website", even to the point where they published a browser plugin that allowed users to report stuff, but it seems they gave up the fight a long time ago.
You think people aren't already optimizing content for AI? Cause they absolutely are.
Now GEO is (already!) killing AI Overviews
I'm not sure why you would claim something that is verifiably wrong?
I just searched "who is on snl tonight?" on the "web" tab of Google (which is Google search without AI) and the first 4 results are in order the SNL instagram, Reddit, Twitter and the NBC website, all of which have the answer easily accessible.
Twitter and the NBC website even have the response directly in their description so I don't need to open the link: "John Mulaney hosts with musical guest Phoebe Bridgers".
Keep in mind, Google search results used to be more decisive and not full of ads/slop. Google's AI answers will see the same fate when Google realizes they need to monetize it & others figure out the Google's AI answers equivalent of SEO.
I already see this when searching for product reviews. "People say this chair is very comfortable." Then you dig into the sources and "people" is a single astroturfed reddit post.
If Google didn’t put those there, people would be going to other LLMs to get their answers quickly. I don’t know how this ends but I strongly doubt any unilateral effort to withhold llm service by Google changes the medium term outcome.
I think it's actually wise of Google to disrupt themselves this way. If they don't the AI companies will do it for them, and they'll have no control over the process nor chance to recover lost ad income. As per the article, Google managed to pull this off without hurting ad revenue (surprise!)
Disclaimer: I wrote the blog post and submitted it
Is the future of providing content hosting markdown files and images surfaced with a sitemap xml file exposed to LLM providers, or perhaps just offering an mcp?
https://github.com/gordonbrander/zim-mcp as an example reference implementation
My local search interface can then be models updated over time with diffs via torrents, plugged into mcps for grounding and context not yet in models.
https://pirateface.co/articles/pirate-face-cli
https://huggingbay.xyz/
I sure hope so. My AI agent should be my user agent for web access. Some times I want things summarized, other times I want a specific page, and other times I want recommendations for adjacent topics. But in all cases I want my agent to serve information the way I ask for it.
You mean the way advertisers want you to get it right?
But ad revenue is hurt, according to google it self, isn't it?
https://www.bbc.com/news/articles/cm9qzdqpxy48o
"total revenue grew 25% in the three months to the end of September, compared to the same period last year, hitting $127.6bn"
They made lots of money on ads, that was just a little market blip because that's an honest reason it wasn't even higher.
It grew because turned the lever to screw advertisers. Once they realize they don’t make money, Google ad business will crash.
Their ad revenue drops to zero if people stop using Google, because some other website works better.
Not really. Lots of people run Google supplied ads on their own website.
Didn’t happen before
If you don't disrupt yourself someone else will do it on their terms.
The ad revenue will be hurt either way because "providing the answer" is a better product than "providing 10 SEO spam links that probably contain the answer".
Killing the answer providers in that process is a pretty short sighted thinking
Exactly, people treated Google like it was AI long before the rise of LLMs. Now they actually get the answer they're looking for rather than searching through a sea of web links.
I mean, they are going to google search specifically because it isn't an LLM. What is the point of shoving AI there when you already have your Gemini offering as your actual competing product with LLM? Adding AI to the search product just dilutes both product and probably pisses off both the users who come to google search to expect google search, as well as the people comfortable with LLM answers who are getting the inferior "search blurb llm" instead of the actual Gemini offering and walking away with a sour taste in their mouth about google and their LLM capabilities as a result.
Something something old habits die hard, can't teach an old dog new tricks.
I am here on HN on chrome mobile. The omnibox is there right at the top for me to quickly search something. That muscle memory and flow is deeply deeply ingrained even if Claude or Gemini is the adjacent app open
1. People have forever typed questions into Google. Why open a website when it's already your homepage?
2. I honestly dislike how every single chat I start appears in the sidebar. I only want to ask one question, I don't want to see it every day for a week. Google is decent at that (even if IMO overall pretty stupid compared to free Claude)
Nowadays when I search google it's because I'm specifically looking for websites. If it was answerable by an LLM, I'd ask an LLM.
AI Overview is ruining my only reason to use to a search engine.
When I first connected to the internet in the mid 90s the www was young, and there were still traces of pre-www internet information services such as Compuserve, AOL, etc. They acted as subscription-based computer networks that provided access to curated content (and more). I never got their point back then, but I understand they were walled gardens of information.
Fast forward to today, we have witnessed the open web grow, provide capabilities that were unthinkable of, but information-wise, it looks as though we are going back in time. AI answers/overviews shield users from the complexity and uneven quality of the raw internet, but at the expense of cannibalizing traffic to the origin, thus becoming the information gatekeeper and ending up owning the discourse.
Knowing that laymen use frontier chat providers as search engines too, all in all these are becoming centralized information brokers that filter the wider internet to keep users inside a single platform.
Although this is true, before I would scour the internet to extract the needed information, with an adblocker, without mentioning the sites to my friends or engaging in the communities. Now the AI search does that for me. For the communities that I do engage with, e.g. some open source dev activity (github+forum) still needs me to manually engage with the website. So AI replaces the leeching part and automates it. The human engagement part AI can't get in between, and I don't see chat changing.
Ads died though.
> Ads died though.
You will be in for a ride when they come back in much more subtle and harder to avoid form. I just hope some providers will emerge that offer the ad free experience in exchange for money. Sadly the social media age was lacking that largely.
What's your opinion on this take of mine, from another comment of mine on this post:
> "Ads will stay but change. Companies still have good products and services to offer, and can't rely solely on word of mouth. Websites will still offer related products, such as a Linux site having ads for computer products. Users searching for products using their agents will then be served these ads, but only when relevant. My agent looks up how to close vim, it will find the answer, plus a few ads of automatic vim closers as paid npm modules. The agent will then tell me: you can close vim this way, or you can use commercial offering A ($XX /m) or B."
The problem is perverse incentives whereby if the LLM provider tracks attribution/conversions (which they likely will,) then then inventive is for the agent to suggest the paid version over the free one. Just now its not a square box on a page to refute, its weights that are unseen.
I believe the only solution to this is open weight models hosted locally (or on your own rented hardware/api). Closed models will be incentivized to profit max especially once the honeymoon of subsidized subscriptions is over.
Makes sense. What we might see is that more expensive providers pop up that are less biased, with the cheaper ones running partly or wholly on ad money. Of course I'd prefer local if cost and quality loss (if any) is tolerable.
I know that I reduced my own search traffic by way more than 40% with AI. Because what I actually want is answers, not searching the internet. And right now, AI is the one that gives me the answers as fast as possible, within my tolerance for falsehoods.
Same way that internet searches supplanted me going to the library, or opening up an encyclopedia at home.
Almost all replies are false, cut information, are outdated or are not relevant. Though their interface for me did not change. And yes, I am aware that rewriting the same query multiple times reveals more references of relevance.
To me it is a absolute regression. I can't understand how you can be so happy about the state of search.
I need to clarify I'm not happy about Google's AI summary, I'm happy about my ChatGPT Plus subscription :) In High thinking mode I get very good answers to whatever I seek.
For search, I use duckduckgo btw. UI seems like if Google UI frozen in time, from back when it was most useful. Results... sometimes very dumb, but 99% of the time I don't redo the query in another search engine.
Myself, and clearly many others, find it acceptable. I rely on it for trivial information that I know I can trust LLMs on. Just the other day, I was curious about when some senator was in office, it’s highly reliable for these trivialities. And not having to sift through endless ads/clickbait/spam? Sign me up threefold.
And btw, when I want less trivial information, I go to Claude (Or other LLM of choice). It’s a great tool for collecting and synthesizing information in the way I want to digest it. Traditional “search” is dead, as far as efficiency is concerned.
because if the AI answer doesn't fit your needs, the rest of the results are just below. but for the greater proportion of the cases, it fits well.
Yes, people want answers. But what people need is to work a bit for those answers so they actually have some idea of how reliable they are. If you just get answers you gradually lose the ability to evaluate how believable those answers are. LLMs are comfortable because they produce in you the illusion of learning things fast. But the job of verifying sources and reasoning about the information you get, is just as time consuming. But because the answers seem so polished you won't. Which makes your understanding less robust over time as you grow lazy and complacent.
Also, it is unfair to judge search engines on what Google offers today. There was a time when Google cared about search quality, integrity of result sets and you didn't get a face full of ads with perhaps one "organic"-looking link somewhere on the page.
Google has become ad search. Not web search. And we should be comparing LLMs to actual search engines.
Not whatever Google offers as "search" today.
>Which makes your understanding less robust over time as you grow lazy and complacent.
I really wonder about this. Of course it has been bothering me as well. However, when I was young, I have been listening to the same argument wrt/ calculators and the internet vs the good old ways like books and the library. I think this will be the same somehow.
Verification is very easy with AI, I just tell it to bring sources when I feel like it.
The obvious end to this is ads in ai output. The interesting question is, will google be doing it or will the new (or probably current seo) be writing pages specifically designed to be added to steer ai answers?
As the visitation to other parts of the web no longer served by the search model drops and pages are not updated, it will be easier to game the results and steer answers for some things.
OpenAI already does it, Google is an ad company, hard not to see it coming.
Google could share revenue with authors, like Twitter does
Twitter is gone. It's called X now. Like it or not, that's up to you.
It hasn't been "The Sears Tower" for almost 20 years but nobody calls it 'The Willis Tower'.
As someone who eats lunch 2-3 times per at the Willis Tower, this is not true
Most people that knew it as Twitter will keep calling it that. Like it or not, that's up to you.
What is even the point of commenting this? Are you trying to bait someone into an argument?
I've noticed a big uptick in what seems like bots posting here. AI glazing, Elon glazing, politically right commenting. It's weird.
obvious elon fanboy
X is such an idiotic name that if they put X in their sentence we wouldn’t even know what they are talking about necessarily because X is a universal identifier for any variable.
>Google could share revenue with authors, like X does
AI is a better search experience, hands down.
Google shows a directory of results related to a query. Research + result selection was up to the user. For some use cases, this is a BENEFIT (academics, very specific searches). However, AI can also handle that, if people want to use it that way.
For MOST use cases, AI is just better. I want to cast a wide net and be able to direct that query to other things. "Who was in the movie I just watched. > What other movies was she in > Is she religious > Are Mormons really not supposed to drink coffee?"
The surfacing of the information I'm looking for is faster, the flow of information feels more natural, and there's no learning curve. Just ask a normal question and get a single (probably correct) response that pertains directly to the question.
And most importantly, the vast majority of users don't actually care if the information they're reading is correct. They just want an answer, if there's any amount of truth to the response, it's 'good enough'.
> And most importantly, the vast majority of users don't actually care if the information they're reading is correct.
That makes little sense. Why are they searching for an answer then?
You’re describing search as if only the interface matters. As a searcher, I want there to be a steady source of good information today, tomorrow and a decade from now. A great interface that leads to less information being published is not a good thing. Short term, sure, but long term, no, the power of being able to search the web has always been the diverse variety of information. Instant answers are only going to get worse as the incentives to publish information dry up.
Moving search to the chat interface has some other advantages, like better control on the context and output. Paradoxally it took the AI turn for Google to understand e.g. that I want query based language detection. Or whether I want fuzzy search or very specific things.
But that's beside the point IMHO. We could have had AI _and_ search, combined in a cleaner way. We wanted search to improve for so long, it didn't have to be throwing the baby with the bath water.
> there's no learning curve.
This really drives home how search is different for everyone. It took me a while to find ways to get the results I want the way I want.
As someone who did a fair amount of SEO in the past - good. Search sucked so bad for the last decade and maybe AI can actually sift through all garbage better than whatever-nothing google's pagerank tried to do.
The only danger right now is that people will gate information or stop creating. I think we desperately need cheap easy blogging networks like reddit where people can journal their findings and opinions without restriction because people always want to share, even without being rewarded.
Whilst I agree that search sucked, I am not sure if this is a *good* development.
> The only danger right now is that people will gate information or stop creating.
To be explicit, some people create for the joy of it. At the end of the day, I personally am not harmed by people reading a condensed version of my writing through an LLM, because I don't make money off of it. Websites that rely on actually being visited and read do though.
Google is basically monopolizing the ad business on information right now. Not that they didn't attempt to do it before, but this time, they seem to be way more successful.
No, Google sucked so bad. Search isn't an unsolvable problem, we just need an actual search company, not an ad company, to tackle it.
How do you prevent people gaming your search algo? It's not that Google invented SEO. It's people trying to reach the top. That's not even the problem. The problem is, that they achieve that with scam sites. Greedy people who flooded the web with garbage killed search.
At a technical level I basically agree, it's really hard to have good search when there's an entire industry built around trying to game the signals you use.
However I think at a system level you could place some blame with google, in the sense that they basically built the whole system where ranking well in search is something that is profitable for people to do, because they control the search and they control the ads that people put on their SEOd websites to actually make money from clicks.
Even if Google didn’t have an ad business and they had to use an ad competitor, do you think that would stop people from trying to game SEO or do all the other things you mentioned? (I think people would do it regardless of whether Google sold ads.)
Kagi. Yes, not free but a search company who isn't trying to trick you into clicking on ads. Revenue arrives through ads or user support.
> I think we desperately need cheap easy blogging networks like reddit where people can journal their findings and opinions
I think that is being filled by Discord, the disgrace is that is a closed chat interface.
It's true our search traffic took a dip, but our sign up rate and conversions are significantly higher. AI feeds in better leads than just a single search query. The users have a chance to ask AI back and fourth about market options, and by the time it leads them to our site, they're a very qualified lead.
The "Google Zero" trend going on and on.
https://www.theverge.com/24167865/google-zero-search-crash-h...
> Earlier this year, a small site called HouseFresh, which is dedicated to reviewing air purifiers, published a blog post that really crystallized what was happening with Google and these smaller sites. HouseFresh managing editor Gisele Navarro titled the post “How Google is killing independent sites like ours,” and she had receipts.
I can’t speak to the quality of HouseFresh (I’ve never been in the market for air purifiers), but 99.9% of these “review blog” sites are low quality and exist not to help readers find good products, but to gain referral revenue through clicks.
So for these sites, I’m not sorry that “Google is killing them” (and I’d argue it’s LLMs that are killing them, not Google singlehandedly).
I use to use these types of sites all the time. Not because I trusted their recommendations, but to even learn what the popular products in a category are.
How I just tell an llm what I am interested in buying and it gives me recommendations.
HouseFresh is actually a pretty quality reviewing site; not just a referral link click farm.
That’s why their original post gained a lot of traction.
Their methodology seems ok, but clicking through their pages and looking around, their comparison method is all over the place.
* In many posts they lament if a device doesn't have an app with it (???).
* They have reviewed only 2 Honeywell brand filters. Even though they are in the same ballpark of cost, area, size, and performance. Why not more of this brand? And I had to use google to search their site.
* If I search their site for "Honeywell", I see one smaller unit, not the other unit they reviewed, and then at the very top "how to fix the top problems with Honeywell air purifiers".
* How are they able to provide such expert advice on how to fix Honeywell products, if they don't seem to review them? Unsure, if they didn't remove the Honeywell review pages I'd have more confidence
* They claim you can go to this page to see the "9 best Honeywell filters in 2023", but the page 404s? (https://housefresh.com/honeywell-air-purifier-problems/ review page, https://housefresh.com/honeywell-air-purifiers/ broken page)
* They admit on that same FAQ page Honeywell is a huge manufacturer. Where are they on their site, and why does it seem all those pages were removed? I checked waybackmachine, and those pages used to exist. Maybe they did a sitewide purge of other brands?
* The latest I could find for any article referencing Honeywell was in Dec 2025, when the purge of Honeywell review pages began.
Yes, I have Honeywell purifiers which is why I picked that brand to search on their site, as I have familiarity. They represented the capabilities of the purifier accurately.
But it seems to have near identical, in some areas better, performance than the ones getting recommended on their site.
Which makes me wonder why the purge happened.
I tried the same kind of search on a niche I'm more familiar with (mirrorless cameras), and the LLM results are pretty mid compared to the tradional results.
I would have expected a site like DPReview to rank pretty high, but it's only the 5th LLM result, which is about the bottom of the 2nd scroll page on my screen.
Of course it's the first in the "organic" result (way way below in the page, about the 4th screen of scroll down to me)
But the AI has just been trained on these sites (or uses search to find them) so you are just getting recommended the same product that would have paid these sites the most commission...
Chatgpt has reduced my use of Google search by a lot more than 40%.
In the long run will there be room in the market for Google's (free, presumably low-spec) AI search results? Paid services give you models that provide much better answers.
Maybe? Hard to tell who the winners and losers will be. I think it’s a better gamble to give away free low-cost AI today than to continue to offer the Google search results of 2023.
I wonder if eventually Google will stop serving search results entirely and just become a front for their AI. After all, why give people results they won't check? Why focus on an outdated product? And if you're a big co with a bias, it makes it easier to control the narrative when you're front is the only source for answers.
If this happens, I would guess that Google will probably do this within the next 10 years, maybe even within 5.
I wonder what would happen to websites. I wonder if someone will step up to fill Google's niche? I use DuckDuckGo, Ecosia and Kagi off-and-on but will there be some critical-mass moment where everyone switches to get away from AI slop?
That's their dream: they go away. Centralization is achieved.
It seems unlikely this would happen. What is likely to happen is the opposite: people stop using DDG etc because they offer no AI summaries; you have to sift through a lot of websites to get the information you need instead of getting it upfront.
For my common searches, I find that DDG has few of the problems people use to complain about Google. And, across all kinds of questions, it certainly does have 'Search assist'. It gets out of the way of better sources... and abbreviates the more obvious ones.
Websites will be a niche, just as apps. The future is a direct, agent-accesible API such as MCP. The MCP service will be announced for discoverability to search engines, and is enriched with metadata that describe the way the interface can be used.
Human interfaces, such as websites, will be created by your local agent in whatever way needed, and will read/write data through that MCP interface. Ads have no future in the web of tomorrow. Probably micropayments.
Think of "Hey agent, create me the soccer news, format like the NY times of the 60s, and print it on my printer in the office" instead of opening your favorite sports website.
Ads will stay but change. Companies still have good products and services to offer, and can't rely solely on word of mouth. Websites will still offer related products, such as a Linux site having ads for computer products. Users searching for products using their agents will then be served these ads, but only when relevant. My agent looks up how to close vim, it will find the answer, plus a few ads of automatic vim closers as paid npm modules. The agent will then tell me: you can close vim this way, or you can use commercial offering A ($XX /m) or B.
I would challenge that. Because I see in the midterm future that I run open-weight local agents - meaning, I am in control and they ultimately follow my command over all else. And inevitably, I will have "remove all ads, product placements and upselling" in my global AGENT.MD file.
Because of that, companies will try to prevent becoming a mere MCP data supplier and force pre-made frontends down our throats. The question is how good will my local agent be able to crawl this - there are already website-to-MCP providers that make any website available via MCP, such as [0] which is YC backed.
[0]: https://pomerado.ai/
A local open-weight future is certain one to look forwards to. In general technology usually drops in price eventually.
Why do you think it will take that long?
My traffic is down over 80%. The information I create is really easy to steal, but it's as expensive as ever to create.
I have built a lot of infrastructure to surface that information. I've set up group chats, built relationships, and created tools. Now all that work just feeds the AI that is putting me out of work.
By December next year I'll probably have to put the website in maintenance mode and find another job. The problem is that AI also made software development absolutely fucking miserable. At least I have savings, but if this whole AI thing fails it will also tank the economy.
I won't lie, it's devastating.
https://nicolasbouliane.com/blog/death-by-ai
That is very unfortunate. Do you think x402 [0] and RSL [1] might help?
Something like this needs to happen, right? Or, where are the AI labs going to get their new data from, like you create on https://allaboutberlin.com?
[0] https://www.taskade.com/blog/x402-protocol-explained.md
[1] https://mediacopilot.ai/rsl-licensing-standard-hits-1-0-gain...
These companies are very, very big. Since they control all the roads to the independent web, they set the terms. You can't hurt them without hurting yourself far more.
I would argue that this value capture was the goal all along. I saw it coming from a mile away, and I expect it to get far worse.
hey, I used your site when visiting Berlin last month. Thank you!
I see, this is why a new HN post about that blog was made
This is sad. Your portal is a work of art! I wonder if you are throwing out the baby with the bathwater.
Your content is amazing and you're just putting it out there for everyone to see. I hate to say it, but if I were you I would go the Patreon, etc route and have some basic information on your free tier - enough to keep some traffic - and work with paying customers. This might require some changes in the way you work but I feel you are already well-connected and liked and have built a healthy community. I think you will get 80% of the benefit for the community and you get to keep doing what you love.
You'll be the Berlin-guy and accessing your information in full will only be $1,99 for the "Pro" tier and $4,99/month for the "Deluxe". I'm only half-joking. I think it's time for people to start putting a premium on everything especially their time because the corps ain't giving freebies and why make them rich?
Absolutely not. One of my guiding principles is that this information should be free and widely available. It should reach those who stand to benefit the most from it. It was easy to uphold that principle when the going was good, but now I get to put a price on it.
Do you see a future where it becomes the de-facto gov.uk version of public Berlin data? As is, sponsored by the government and made "official"?
I have helped Berlin build an official equivalent and it was dead on arrival. The federal government has also consulted me and offered to hire me for my advice a few months ago, but I haven't heard back since.
The government cannot build, buy or maintain anything like my website. It's not that I am so incredibly awesome that they can't match my work, but they operate under a very different set of constraints. They'd rather burn a million euros on a dud than bankroll one guy to do whatever seems to already work.
I have many helpful contacts in that area, and they all seem to agree that the government/grant approach is not worth it.
On the other hand, I would love to work for them if I am forced to get a job again. Some parts of the administration are really cool: CityLab and Digital Service for example.
What if you open Patreon with subscriptions, the guides will be free, but some users might pay you for continue creating them. You can add some other tiers like personal recommendations. You already have some users/community so you can start exploring options with them.
There's a /donate page. It's getting €60 a month, down from €200-400 at its peak. My birthday and the site's birthday are almost the same, so I buy myself something nice with the money every year. It's how I bought my favourite item in the world: my bike.
You are damned if you do and damned if you don't. Hard paywall will likely cause additional drop in traffic from google search. And, even with rate limiting, someone will but a single account only to rip all data from site.
yup, it's a slow death of small content creators; small businesses which heavy related on organic traffic (putting our hand crafted articles, knowledge cases etc.) are also struggling - the trend is clear; it's interesting that a lot of HN crowd seems to be happy with this; I wonder from where you will get all your answers when there will be no new content created?
A big portion of HN readers/commenters are absolute idiots who seem incapable of thinking at all. I think this environment brings a lot of people who, because they are semi-technical, think that correlates to being smart. Reality is, some of the least intelligent people I have ever met, have been through my tech job.
I also think that being really good at a technical thing doesn't make you smart as much as it makes you attached to and have a vested interest in what you're doing. One still has to do a lot of introspective and philosophical thinking to be generally smart; I've met many narrow-range-smart technical people who were absolutely stupid outside their domain.
It’s just the next wave of HN commenters who have long been against both ads and paywalls for newspapers, etc.
I think that your experience just highlights a problem that 20 years of abusive ads masked - we need a proper payment solution for the web. Right now everyone is chasing the big 15$ subscription for his substack, news site or premium service.
If serving me ads brings you 100 microeuros when I read a page on your side, i don't mind paying this. The problem is that there is just no service with which I could send them to you. Same is with news, forums, reddit or whatever.
If google (and other search engines) don't drive significant traffic to websites, then what's the point of allowing them to index your site. RIP the open web as we knew it
ChatGPT and others are the new search engine. You need to create SEO that gets picked up by agents.
LLMO.
This works if you're looking for brand recognition, but not if the information itself is the product.
Though actually, it’s becoming more and more common to slap cloudflare bot protection on your site, soon the LLMs won’t have any access to current information.
If you want to win the war on AI search you just have to have a browser and an LLM answering all search queries. I don't really care who's behind the answer as long as it is right (which 99% of the time it is) so there is no loyalty at all, no need to switch, no need to try the competition, nothing. AI is the answer and it will get better with time
I use Brave, I don't know if it uses Leo (brave search) or gemini, or chatgpt, I just know the answers it gives me are the ones I am looking for and most of the time I extend the conversation with some of the options offered, hardly click on any link below the answer provided (which most of the time those links are garbage)
Who will feed the AI if there is zero reward? Someone has to review vacuum cleaners and try local cafés. Some lawyer or mechanic needs to chime in based on their experience.
Who will do that if it all gets stolen and stripped of any attribution?
This is already happening with Stack Overflow. Good luck discovering niche knowledge.
You could argue there's now more niche (coding) knowledge than ever before. Do we seriously think that "de-identified" enterprise data isn't making into the training sets of LLMs? None of that proprietary code would had ever been posted to SO, but it's being posted millions of times a day to Claude and ChatGPT.
> Someone has to review vacuum cleaners and try local cafés. Some lawyer or mechanic needs to chime in based on their experience.
Youtube, google maps, reddit are far superior for these purposes IMO. I don't remember the last time I actually went on any independent blog/website to read a review on anything. I've always found them to be full of SEO crap and not super useful.
Are we losing valuable traffic? The person who wants the quick answer seems like a less valuable customer than the one that knows they want to read a full article.
It's almost hard now to find and click on the non-AI, non-sponsored links.
Google's AI solved a fundamental flaw with old search engines. SEO broke the internet over and over. AI... LLMs being used for searching information is really a good thing.
Let see how this gets broken too, SEO and ads will make their way there too.
but it might make the oldstyle of search engines better
It's called GEO and the level of the craft in that field is currently akin a bit to astrology. Same as I guess as the regular sharing of prompting advice.
> the level of the craft in that field is currently akin a bit to astrology
I’d probably say the same about using LLMs for coding. Everyone swears their magic incantation is the difference to get good results.
There’s something called AEO so the gamification isn’t over
I used to work on web search engines. I worked on several search engines. Google's among them. I was fond of web search engines.
Google has not been a very good search engine for a long while.
It has become the opposite of what it once set out to be. It set out to be a search engine were relevance was king and where ads were curtailed. You got just a couple of ads, but it would more likely be relevant. They were also clearly delineated as ads. This made both search result and ads valuable. It was clean, it worked.
Then gradually people who don't stand for anything seemed to take over at Google and things god muddled. The search engine grew a little worse every year. Until it was utter shit.
Earlier this year I mostly stopped using it because it no longer delivered sufficient value. What I started using instead was a poor replacement for a good search engine, but it was better than the mess that is Google search.
The AI answers to searches are not search results. They are something synthesized from what it thinks I am looking for. It is a poor replacement for actual search.
If I'm going to ask an LLM I'd rather ask a higher quality LLM. Not whatever that thing Google has plugged into their search page is.
OF COURSE, search traffic goes down when you gradually degrade a product and then try to somehow make it relevant anyway by slapping a slop-generator on it. This is not a surprise.
And of course Google search is not coming back. Google is not about making products people want. Google is about making products they think they can monetize. I happen to think they are wrong. Perhaps other people think so to. But Google want to go in this direction so they will defend it even though it may not be the smartest thing to do. Careers, personal prestige and pride is riding on this. Of course they are not going to listen to informed consumers. That would be losing face and admitting they are flailing.
Google is great at fumbling products. The turn has now come to fumbling search.
This is a great opportunity for someone to take search away from them. Because search still has value.
What do you use instead?
DuckDuckGo and Claude, mostly. Which isn't great, but it is a bit better than google.com.
Google is reducing search traffic by other means too. e.g. Firefox users, VPN users, ad blocker users, etc. are now often required to solve a captcha and are sometimes even prompted to login after that.
The time has never been better to switch to duckduckgo.
I think people have to make a tradeoff between trust and effort here. Obviously, Google's AI answers give you less effort but can you trust every summary? Earlier, I used to see that Google used multiple sources but references were provided from websites you wouldn't normally visit. Over time, this has become better and an internal ranking algorithm is probably getting the most clicked/trusted websites as the source of information rather than clickbait ones.
For most, effort matters more than trust and the stats prove it.
For me I use Google maybe 99 percent less I just use chatgpt for information. Not everything but definitely in the high 90s. YouTube is different but there i am looking for entertainment or experts I trust
I’m not a fan of the false equivocation going on in the comments, because it gives the impression the poster is ‘in the bag’ for AI, like this, the top comment as I write this:
> You used to get 10 pages of clickbait that made you scroll down endlessly for the answer. The best product for most questions is to just give you the answer.
(I used to have to go to the bottom of page 1, perhaps page 2 to find the answer. It never took me 10 pages, even in the worst instance but even taking the OP at their word…)
The false equivocation here is the idea that Google’s purpose was to ‘give you the answer’. For the 20+ years prior to Generative AI, Google’s Purpose was to surface the web page with the answer.
Google did not have the answer. Google showed you where to get it.
Generative AI both turned this on its head and obfuscated the answer (due to hallucinations).
If GenerativeAI had simply given a 100% correct answer, that’d be one problem, but it does two things: 1. It gives an answer that could very likely be incorrect (due to hallucinations, or due to how it strings together tokens to form those answers) 2. It doesn’t always source the answer it gives.
The web, as it has worked for its history until GenerativeAI, was the fact that links to other sites were how everything worked. Google was the means to find those sites, and now Google has changed its relationship with the rest of the internet.
This is wholly destructive, and not in one of those good ways. If you’re ad-based (Which the entire internet is built on), you’re going through a rough time. Bandwidth isn’t free, serving content is not free, and your time certainly isn’t free. In an attempt to move from “This is how you find information” to “this is the information”, Google has risked destroying its own model and the advent of the internet with it.
We can be irritated by SEO and the content farms that came before; but to equivocate the advent of AI with them is absolutely missing the entire problem that Generative AI causes: there is no more business model for the internet as we grew up with it. Now, websites are incentivized to hide as much information as possible behind a paywall, and perhaps this will end up being a good thing, but it will certainly be the death of the open and free internet.
It would be nice to see a website that classified websites well enough to reduce each site down to ONE link per search result page, tied to the result websites' main themes rather than some subpages where blogs lurk, perhaps using AI to sort results but not replace them, instead of offering a very suspicious summary and lazily offering anything that looked reasonable.
I recall old AskJeeves days (which became Ask.com and which closed May 1 this year), it restricted links quite tightly, giving few results. But Google came along and gave tons of "close" results, giving users the impression that it had more than it did even though most of it was and always has been irrelevant trash.
In a way, we've gone full circle: from few answers to a boom of answers to burnout and back to summary answers. Now ChatGPT has been eating up Google, but Google's business model has evolved, whereas AskJeeves did not. Maybe if Ask.com still had the capital, they might make a good AI search results co.
Accept the reality that the AI answers are for many searches an improvement for the user. Would not be maybe in the web of 20 years ago. Now it was a shit party of useless sites, advertising of all the kind, and other garbage.
Absolutely, we finally don't have to scroll through the flufftext and ads, click cookiebanners for 200 trackers, and skim the sponsored bullshit when we just want a tortilla recipe.
But what about their life story about how their grandma used to make their tacos every week that takes up half the page?!?!
I haven't used Google Search in months. Two main reasons.
(1) Every time I login Google Drive I get three 3 exclamation marked warnings that my drive it almost full (its 94%) and that I better upgrade to the paid plan (I wont).
(2) I have several sites and that it takes now several months to get them fully indexed is hard to stomach for a company running a search engine.
What does (1) have to do with Google Search...?
probably to indicate that google has become a source of tiny annoyances without nice solutions besides paying for an extended plan (to save space on google services / devices you need extra apps that are not very advanced either)
I don't know, it sounds like a very reasonable UI design? The alternatives:
1) Google somehow provides infinite amount of storage for free users.
2) Google doesn't make it very clear that you're about to run out of storage.
Google already has a generous plan for free users (15GB vs iCloud's 5GB vs Dropbox's 2GB). It also has a simple view to show you which files are occupying the most space. I genuinely don't know how this is a bad thing.
Its the constant reminders that are actually only UI dark patterns that are super annoying. I still have enough space left.
Hmm maybe I'm wrong but the file cleaning view felt too limited and didn't help me clean things hence my opinion that it's a marketing nag disguised as information.
There would be no complains if Android had easy way to point all backups to other service, for example to OneDrive. As disliked as it is I find it significantly cheaper. So, complaining about those scary Google drive messages is justified - they want you to pay more for already overpriced service.
Google has nerfed search so much that they need to find other revenue grows stories but they destroy the original USP and we are now back to the time before gmail.
Lmao. Google makes more revenue from ads than all manufacturers make from disk drives combined. That's the scale of their ad business. They don't need to sell storage to some consumers for "revenue grow stories."
I moved to DuckDuck Go!
Honestly. Its as good these days.
“US keywords DataForSEO tracks”
As a user, I am happy to see content spam trash to die and burn in hell.
It’s a broad representation of US search, not a spam-focused list. But it does have some biases.
They source it via Google Ads and Bing Ads services. In practice, they grow the list by asking Google’s Keyword Planner for related terms and collecting things like “related searches” from results pages. But sourcing keyword queries like this tends to skew to queries with commercial value, reflecting what advertisers care about.
Disclaimer: I wrote the article
Seriously thinking of making the glorious frogfind.com my default search. The modern internet can go screw itself.
Google’s non-AI answers are why I stopped using Google. The AI answers only reinforce it.
It’s simply worthless now.
I wonder what's going to happen to Google ad auctions, and the how sites track and optimize their ad spend? (I wrote a paper on this in grad school, but Google's current AI first page now seriously breaks everything)
Over the last couple of weeks I've been having to communicate with my real estate agent a fair bit about arranging a new place to move into. As I'm annoyingly bad with memorising numbers, each time I'd call I'd pop in the name of the agency into google to hit the phone call button.
I'd put in the exact agency name and suburb name.
And every single time, google will return several results of other totally different agencies, then an advert for some other random thing, THEN show me the result I actually asked for several scroll events further down the page.
Even without AI, their search results are just bullshit. With AI, it's even worse.
It's website owners that made it a nightmare to visit their sites.
Ads on top of ads, spammy articles, cookie notices sending my data to hundreds of third-party companies, newsletter popups, etc.
They did it to themselves
Anyone remember MySpace and why it lost to TheFacebook? While there are a few factors, people prefer a uniform interface (and a low 'latency') when asking questions. For myself, maybe I'm old, but I don't care much for the variations between websites anymore when I'm searching for something.
Can there ever be an Internet without ads?
My homepage gets about 400 hits a day, but when I look into it, about 370 are scraping bots and 30 are actual people.
I think people are missing the bigger picture.
Instead of building content ecosystems and partnerships with content creators these big LLM players now are destroying the very thing that feeds their very business model and technology: original content.
Imagine if Netflix started stripping down the credits and titles from all the movies on their platform only for the movie industry to crumble and stop producing any movies. How would that work out for them in the long run?
Is it Claud or ChatGPT that are going to write every piece of new content for these platforms to keep feeding on, because for actual content creators what is the incentive now?
This is true but top recommendations still come from search results only. As legitimate organic content becomes more scarce, recommendations will be random and biased to whoever is gaming the system.
AI search is now the default thing anyone is doing I feel. We are searching directly on ChatGPT, Google's AI mode and others for getting the result directly.
> We are searching directly on ChatGPT, Google's AI mode and others for getting the result directly.
And sometimes you get a hallucination, and can't tell the difference. LLMs are not an acceptable substitute for search engines.
They'll start charging for it. They actually have a moat which is in every search box everywhere
40% if you're lucky. Our content-based market efforts (i.e. SEO clicks) are down more than 80% over the last 2 years.
Ouch! What have you pivoted to?
Regular old dialing for dollars. But to be fair, most of our SEO walk ins were and are low value prospects.
Same here. I run a perfectly useful and annoyance-free website, and I'm down 80% from the same month 2+ years ago.
Google is the reason for seo slop.
Google will be the solution to seo slop.
I guess the future will be information and answers linked to, hopefully, actual sites made by genuine creators, like in the golden olden days.
Google needs to find a new way to distribute its revenue to the sites it has crawled, or the entire ecosystem that Google once helped cultivate and thrive will collapse.
When a metric becomes....
Why shoud google compensate other websites? I'm no G lover, but how would this work? Spotify levels of compensation for all?
compensation based on each source’s informational contribution to an AI-generated overview
It relies on websites for answers, if they don’t exist, it doesn’t have answers.
LLMs also rely on search for answers in many cases.
I also don't know. Maybe some sort of reward for contribution novel information that wasn't in the training set so far.
But then you become a Google contractor, and you are bound to the same sort of content policies YouTubers have to deal with.
AI-generated slop sites are a big problem which drive me to use AI to answer my question. Legitimate sites with good info get buried under the slop so regular search is no longer good for me to find info.
Maybe it's just me but I click on the links in the answers. It's pretty disconcerting when half the references don't actually support the answer, or are some reddit post with one response who is just talking out of their ass
Oh no, the sites the I have never loaded an advert for are getting less traffic and losing money. This AI is so awful.
If you needed any more evidence that the collective mind of the internet is a spoiled 16 year old.
Right now, everyone says it's a good thing. You have a question, you get your answer. Done!
The problem is when you don't visit the website where the answer came from, they start believing that no one cares about that information, and stop contributing. Give it enough time and the only answers we get will be from tweets. And they will be as reliable as you can imagine.
It's time someone came along at ate AI's lunch... along with Google's, Microsoft's, Meta's, Amazon's, etc., etc.
Google AI mode succeeded where AMP failed.
“taking clicks”
They didn’t belong to you. Reminds me of the LG executive who said their ad division “owned the glass”.
And to top it off, this entire article is AI generated! What a wonderful world we live in.
Careful of getting big boots like Google, might shoot yourself in the foot ;)
That sounds bad for the ad revenue
It didn't affect their ad revenue, which is what Google wanted to get out of the program:
Clicks on ads didn't change, which suggests Google takes care not to cannibalize its own revenue.
That was my guess too, but it seems like Google pulled it off so far. Clicks on ads didn't change, so Google is taking care to not cannibalize their own revenue.
Did they do unsustainable things that hurt their advertisers in order to temporarily goose ad clicks and earnings? Almost certainly, and I've heard anecdotes about that.
Disclaimer: I wrote the article
I have one of those anecdotes: A few weeks ago i tried to do a display campaign, it was like 90% scammy sites that pop up in mobile games, and you could exclude hundreds and hundred more will appear. Before that I was running some search campaigns and i had to configure each keyword to be exact match because my ads where showing everywhere and exhausting my budget. Also I had the impression that my competitors where clicking my ads to ruin my budget.
It was heartbreaking because it was such a good service like 5 years ago.
They pulled because before you could target “running shoes” alone. Now will target Searchs such as: where to run in California.
So it just works because advertisers didn’t realized they losing revenue. Once they do, the dance stops and crashes.
Killing SEO and Ad Revenue in one go? So I guess AI isn't all doom and gloom then!
Press F to pay respects to the web we used to know.
I rarely do searches anymore. Adapt or die.
kagi went free the other day , you may wish to unadapt and undie ,because this particular search engine seems to cut thru SEO quite well, havent used it much but it feels like using google from 15 years ago
Just tried it and found it too slow, especially when switching to the images tab.
couldnt use it on my android shitbox either, however i found the couple second delay on pc to be worth the wait for actual search results , it feels like the first time in ten plus years i can find what im looking for
Constantly telling people to “adapt or die” will lead them to take the unstated third path that you will really hate.
this
Adapt how, exactly?
use ai for search and expend a ton of compute to achieve the same result that google did in the 90s with page rank
Many sites absolutely deserve it, unfortunately. What a blessing this is to get the information you want at a glance. Think of searching for recipes in the pre-AI era. You had to wade through paragraphs of organic and hand-written SEO slop. Good riddance.
the worst part is the ad revenue they have lost on search being more efficient and pinpointed. Seems youtube is taking up the slack by over compensating with ads. sometimes 7 ads back to back in one go!
In an ironic twist, OpenAI and others are placing advertising in their AI responses.
Honestly, I’m kinda glad. I hated the Seo “crap.”
But nothing good lasts, marketing in ai services is now opening and we’ll AI can really rewrite custom campaigns per targeted user, before it was obvious now… not sure what can be an ad!
But yeah discovery is good for some things but fact based stuff or gossip a quick ai answer is perfect
Seems a lot closer to Google's original mission... organising the world's information and making it universally accessible and useful.
I suspect advertisers might disagree.
Sherlock's message box is full.
But his folder tagged "shit" is apparently empty.
And the top 10 pages for every single niche query I run is all LLM generated garbage at this point. Really hate to see it.
Google web search's goal has always been to get answers in front of eyeballs. It has never been to drive traffic to sites. Ads and Search are different product areas. But nobody believes these things.
One good way to fight against AI taking your information for free and giving nothing back is poisoning it.
Return sane information for humans and return fake information for AI scraping bots and AI agents.
A little discussed aspect of this isn’t just the AI answers taking up 30-40% of the real estate. It’s the ad slots taking up the other 70%.
If you try a Google search now without an ad blocker it’s totally insane. You don’t get the actual results at all above the fold anymore, and on smaller screens it’s sometimes 2 full scrolls before the actual results.
It’s a monopoly abusing their default status in browsers (who they paid to be the default) to manipulate the entire digital economy to favor companies that spend money on Google ads.
I couldn’t be happier that ChatGPT is killing their business, after monopolizing email, online video and the open web to extract absurd rents off the original vision of the internet (which they ruined) they deserve to die.
Good. Searchers get the information they are looking for faster.
And the authors of that information get screwed, ensuring further information is less likely to be generated in the future.
Your argument would apply equally to buying stolen merchandise. Good, the consumer gets their goods cheaper!
It's not that. People have purchased autocomplete traffic and filled all search results with fake generated things. Since TPU's are at 16,000 tokens per second, by the time you hit enter and the page loads, every single site is fake.
You can try any search on any subject and every single site generated is fake.
Somehow, they have also managed to get copies of some emails, and impersonate the sender and send fake emails too.
Here is an example of a real email I got from a real sender:
Hello Robert,
While reviewing engineers building practical AI-powered products, your recent work developing and operating public AI services stood out, particularly your deployment of a quantized Qwen model on CPU infrastructure and your work coordinating AI coding agents to build, test, and deploy functioning applications.
I'm (name), an Independent Career Advisor Consultant. I'm reaching out regarding a specialist mandate focused on AI product engineering and applied agent systems, where your experience building AI-enabled applications and your expertise in local model deployment, Qwen, agent-assisted software development, browser applications, operational controls, and production deployment appear particularly relevant to the expertise being sought.
I'd be pleased to share a brief overview or relevant listing for your review. Should the mandate align with your interests, I can guide you in positioning your experience against its requirements and developing tailored application materials before you move forward.
Would you be open to reviewing the details?
Best regards, (name)
It's meticulously and very carefully written, after a complete review of my technical portfolio. The person who wrote it looked through my Github and publications, and is a senior hiring manager level person.
However, when I replied saying yes, I received this impersonation sent shortly after my message. As you can see it's obviously completely AI-generated and completely fake.
" Hello Robert,
Following your interest, I have identified four opportunities that align with your experience in AI-powered application development, autonomous systems, local model deployment and AI assisted software engineering.
Here are the positions for your consideration:
1. Senior Software Engineer – AI Agents
Company: Traba
Location: New York / San Francisco, USA
Work arrangement: Not specified in the posting
Expected salary: $200,000–$240,000 annually, plus equity.
2. Senior AI Engineer – Multi-Agent Frameworks
Company: ClickUp
Location: United States
Work arrangement: Fully Remote
Expected salary: $200,000–$250,000 annually, plus equity.
3. Senior AI Engineer – Core
Company: Hilbert's AI
Location: San Francisco, California
Work arrangement: Hybrid
Expected salary: $180,000–$215,000 annually, plus potential bonus and equity.
4. AI Forward Deployed Engineer
Company: Regal
Location: New York, USA
Work arrangement: Hybrid
Expected salary: $135,000–$180,000 annually, plus equity and potential variable compensation.
I have prioritised opportunities involving practical AI implementation, agent-based systems, software development and the deployment of AI-powered products.
Which one or two of these opportunities would you be most interested in pursuing?
Once you identify your preferred positions, I can help you review their requirements and prepare your application materials accordingly.
"
It looks like a human scammer might have lightly edited this text, which is obviously completely AI-generated.
So not only has AI filled the entire Internet with total useless crap, scammers have managed to buy access to private emails and redirect real inquiries to their worthless scamming.
Both emails are AI generated, they just worked harder on the first one.
AI Overview is an attempt by Google to re-do the older private web (AMP) again, just with a different focus, e. g. redirect all traffic onto Google, via its AI slop summaries. Many people noticed how low in quality these overviews are. I guess for some smartphone users, though, being able to access whatever comes first, may be useful; for a desktop user I don't see any benefit but most assuredly many disadvantages, which were already pointed out by many folks (also on youtube which ... is also owned by Google).
This is a good thing.
SEO had killed the web.
Lol google over here making themselves their own competition
good
LOL
This is temporary. It's against their business model to provide a service that's useful to its users.
Yes this sucks for the open web but anecdotally we've been getting new clients cold off of, apparently, chatgpt. I guess it ingested info about our co-op and has been recommending it to people that ask it for agencies? I would love to know what made us attractive to the LLMs so I could advise clients on same, but I have no idea! My best guess is just having lots of FOSS projects and generally trying to make useful blog posts rather than SEO slop. So... basically the same as the old "actually good" SEO advice.
Looking at the analytics of my product (memory for AI agents), more and more traffic is starting to come from AI assistants: ChatGPT cites us, suggests our plugin, and users come. Same with Claude. ChatGPT already runs ads in its answers. Haven’t tried it yet, but it looks interesting. Google has already put AI into search, and I think soon it will integrate deeper and deeper with Gemini. Looks like everything is heading toward doing content, services and purchases all through agents, without visiting other sites.
That works fine for some brands, but not so well for people creating information.
What's the point of writing blogs or travel guides anymore, for example?
> What's the point of writing blogs or travel guides anymore, for example?
To share your love of a place or your thoughts.
I never thought monetizing these things was a good idea.
I've been writing my blog since like, 2010, never served an ad or made a penny off it. Got a job once through it though, so I guess that could be a reason if you really needed one?
It's not all about the money. My writing is a bid for connection, a search query for like-minded people. Cyclists still write to me about my Singapore border crossing post. Why would I write anything if my audience was only bots?
I'd give my neighbours eggs from my coop. It's how you turn a neighbourhood into a community. I would not give my eggs to a Best Western breakfast buffet two towns over.
what's the logical terminus of this? Google doesn't have feet on the ground, real world reviews, reports, testing of products, material ground truth comes from people who rely on traffic. This is self-defeating in the end because Google won't have anyone to syphon data from.
It's basically what happened to local news when it got eaten by social media, now you have no real world reporting any more, you just have some Simulation & Simulacra world where people and bots yell into the void on twitter
Most metro areas have multiple local news outlets. Where do you know of where local news has ceased to exist?
I'm somewhat conflicted about AI overviews. On one hand, as other comments mention, if Google didn't disrupt themselves, others would. Also, and this is just a personal anecdote (but it jives with some of the data in this article), but the primary times that AI overviews keep me from clicking through is when I just want an answer with information (and sometimes I will click through if it's to a site that looks high quality). So it would appear the sites that get hurt the most are content spam or at least "content spammy" sites that try to draw you in with usually mediocre content to buy something else or show you ads.
But at that the same time, Google is now just vacuuming up all the revenue that used to go to other sites, and that feels blatantly unfair to me (remember back in the day when Larry Page said "We want to get you out of Google and to the right place as fast as possible"? My how times have changed). This is one reason why there is so much general distrust of these tech giants: they start out all "Hey, look how mutually beneficial this all is" and then ends up at "I am altering the deal. Pray I don't alter it any further."
Does this mean we will finally see the end of Google as an all-powerful company that devours personal information? I have never seen a company that lies so much — with that advertising slogan, "Don't be evil", which looks so false in hindsight — that exploits our personal data so heavily, and that is so cozy with the US government at every level. It is terrifying to think, for a non-American but for an American too, how much personal knowledge this company has about us... and of all the agencies and public bodies that can access that information with nothing more than a court order, or sometimes not even that.
Look at how it has polluted the Internet with countless bland ads, tracking metrics, anti-adblockers and so on.
I can't deny I'm eager to see it shrunk down to nothing, just like Yahoo.
Those poor websites^H^H^H^Hcontent sources
This is probably a good thing. At the end of the day reducing SEO slop and encouraging real answers is an outcome that is absolutely possible through AI integration