There's going to be a short and tight window for the non-dominant search engine companies out there to seize this opportunity and make their search results based on real things. AI search results have already shown to be hallucinating on basic queries.
When I look up something as basic as time zones in two different regions, I don't want AI to imagine something, I just want the results from the thousands of sites available out there to just be presented to me.
But Google has shaped the internet to be the way it is. At some point over the last 10 years they gave up on fighting inorganic seo spam sites and decided to join in and profit off of it.
Often asserted by people who sound very confident.
I have a Gmail account that I use only for purchases. A majority of the mail there is about me spending money. A highly valuable consumer, you'd think?
A far as I can tell, nothing about that account has been passed to any third party, for money or for free.
I’ve read that page, and I see the part where a third-party will upload their own data to google so that google can target ads more precisely; but I missed the part where a third-party gets access to a user’s data
It's not 2004, Google has more ways beyond just Gmail to hoover up behavioral data to feed its ad platform. There's Android OS, Chrome, Maps, Youtube and Google Analytics. You live in a panopticon, you just don't know it.
Not really. AI overview is a natural progression. It is pretty accurate. The hallucinations have decreased significantly over the past year. Most people won't mind this.
I've had access to this in beta for a few weeks now and contrary to a lot of the very vocal opposition here in this comments, I have had good results.
If you're searching for a specific site, obviously this won't help, but I would imagine that an increasing percentage, if not the majority, of searches are people just typing questions into the search box. I certainly do that a lot.
This gives you an answer right at the top, and I haven't seen any obvious problems with hallucination yet.
I think we know exactly how Google will monetize this, assuming they keep their search monopoly. Over time the answers will get increasingly skewed towards their paying customers. Eventually Google's endless need to increase revenue at the expense of their product will render the results quite bad and everyone will start looking forward to the next paradigm shift.
Can you point to any evidence that Google intentionally skews organic results towards paying customers (as opposed to, Google's paying customers have the money to afford expensive SEO to get themselves a higher ranking)?
1. Google intentionally makes search results worse so that users will do more searches and see/click more ads. There was a big thread about this somewhat recently.
2. Ads have steadly taken up more of the results page and become less distinguishable from organic results. This causes users to more often click on whoever paid Google more, rather than what Google thinks is the most relevant result.
AdWords and AdSense revenue goes to them instead of going to "content creators". Which is pretty good for google.
If you think that's bad, just wait until they can use generative video to create reviews of new tech products. (Come to think of it, just wait until they can AI generate video reviews of any products.) A whole swath of people and organizations relying on youtube revenue are going to be taking a 95% chop in pay as Google, or whoever wins generative video, takes all of the youtube payouts.
We come to the inevitable problem, the problem that has been obvious to everyone from day one. If you cut out the content creators, what feeds the AI? Where does it get its information from?
I think people haven't fully clocked yet how foundational this shift is for the wider internet.
For decades now, we had a social contract: Google scrapes our websites and surfaces some of our content, we make it easier for them, Google gets people on our websites, and in exchange Google gets content (links) to stuff with ads (plus they get some tracking on top).
This created an ecosystem where people are incentivised to create better websites. You create good content, Google surfaces it, you get eyeballs, and it's a decent deal overall.
Sure, this dynamic is contingent on Google fighting SEO spam, which they seemingly stopped doing lately, but still there was hope that it's temporary. There was hope that if search gets bad enough they'll start bleeding users to competitors and get their stuff together.
Now Google throws the contract out of the window. Whatever you write or record or snap or shoot, they ingest everything (because we have no choice) and feed it into their Moloch, and out goes unattributed, unverifiable, ad-ridden slop. There is no compensation, no recompense.
I think it's even worse than that. I think we'll see a shift in incentives for writing content. If websites are used as the source for AI engines that people rely on, those websites are effectively becoming the source of truth for everything.
Right now I can tell if a website is blog spam SEO garbage, but I'm not sure if that'll be possible once everything is "laundered" into a homogeneous looking result.
I think there's going to be incentive to create huge quantities of "data" because having the highest volume makes you the source of truth for a topic. So, rich companies and people will be buying compute like crazy so they can generate tons of AI spam so another AI surfaces their viewpoint as "fact".
Ugh yeah, I hadn't even realized what a serious vulnerability it will be when large corporations and foreign governments start to flood the web with their preferred "facts". Google and Microsoft will gladly slurp up all of that training data. Modern Google doesn't even bother to exclude blatant spam domains from SERPs, there's no way they are going to combat this problem effectively.
This really is looking more and more like the death of the open web. Curated and closed communities will be the only places left to get authentic information online.
Do the AI features apply to content from sites that block their 'Google-Extended' crawler (which they use to index content for their LLMs)?[1]
I'm not as worried if that distinction still applies – I can just block 'Google-Extended' and let the normal Googlebot through. If not, though, that feels like the beginning of the end of people wanting to be indexed, eventually leading to Google eating itself.
I have felt for about six months that the web as we've known it is dead. The contract you speak of is on its last legs.
The problem is, with GPT style AI I think this was inevitable, Google has just decided to ride the wave rather than falling into obscurity.
Also, google has youtube, which should not be underestimated as a source of information. There is a lot of good, credible information on youtube, with even better signals (sub counts) than the traditional web.
This is Googles ace in the hole so to speak, the web can die, they can still feed their AI's with youtube, who else can?
> It’s using a new specialized Gemini model to summarize the web and show you an answer. It’s even using Gemini to design and populate the results page.
This sounds like a typical overfitting for simple, generic searches, at the expense of anybody who needs precise results. Since google started ignoring operators in some searches, targetted searches have become more and more difficult.
It's almost impossible nowadays to search some random error message you encounter from an open source library while coding, even if the string is present in a relatively popular library on GitHub. I've been forced to use Bing for my error message searches now.
Targeted searches have become almost impossible anyway. I was trying to help my dad fix a fridge a couple of weeks ago and searching for info was infuriating.
The problem I had was the I'd enter the fridge's model number as part of the search and Google knows it's a fridge, so instead of "exact model disassembly" I'd get results closer to "fridge disassembly". It was useless.
And I felt like I was getting crushed by what I call synonym syndrome. When searching for info about the condensers, the search goes from "exact model condenser not cold" to "fridge condenser not cold" to "refrigerator condenser not cold" to "refrigeration condenser not cold" and suddenly I'm getting results for vehicle air conditioning problems.
I've never had this problem when searching for an exact model number. You can find a couple of sites that have the manual in the first results; the only trick is finding the one with the least number of ads/lowest sketchiness.
starting from https://www.repairclinic.com/ is a great entry point to chasing down repair information. Searching directly in YouTube too can give exact model number matches at times.
Doesn't an AI search engine cost more in terms of energy? I would also wonder how it avoids being tainted by a web of other AI results of various qualities that it is indexing. It would appear to become more expensive to operate, thus requiring more ad revenue to support it but providing less useful results over time.
> Doesn't an AI search engine cost more in terms of energy?
I'm pretty sure it does. I'm convinced this AI bubble will pop when it turns out that many of these use cases are not actually profitable. For now there's still plenty of cash left from zero interest era to throw at subsidizing this stuff.
>I would also wonder how it avoids being tainted by a web of other AI results of various qualities that it is indexing.
I also wonder that....how Gemini and other LLM chat bots decide what to index and include in an answer and what to not include. I guess good old PageRank[0] would be decent here.
Oh good, now I can get AI generated spam results directly rather than having to click through to the SEO spam pages that Google loves to serve.
More broadly though, it feels like this is going to accelerate the decline of the open web. Fewer and fewer people are going to create quality content as Google steers more traffic towards their own generated content derived from other's original works.
It was discussed a number of times here on HN a couple of years ago and the author is an active participant here.
Please note that it isn't a very large index, but it shows what one single person can do today on spare time and until recently a hobbyist budget.
It also very effectively shows that every dumb excuse Google tell us for putting spam on top and not respecting our queries are just that: dumb excuses.
As far as my observations go Google search results has been manipulated by something AI like for a decade already and part of the reason why I pay for another search engine is to get away from that.
No, I am not talking only about external SEO where some company pays to create links to their page using specific keywords:
This also happens with articles about minor Javascript frameworks that definitely doesn't pay for SEO.
What happens is I search for some less popular (today) framework that a previous dev included, and I get s
what seems like interesting results, but when I look into them it turns out many of them are about a completely different, more popular(today) framework.
Undoubtedly someone has put their soul into creating such a system and it is somewhat impressive - but also absolutely terrible when I try to find something and I end up doing unpaid QA work for one of the worlds biggest corporations.
Which is why I pay for Kagi. Yes they use Google and Bing behind the scenes but somehow they have managed to get rid of these massively annoyingly false positives that has destroyed my relationship with Google over the last decade.
I had their AI thing on for a few weeks and it was the most annoying thing. I want a quick google search and response, not waiting on a block of text to be written then the entire page of results shifting down.
And now I can't even use google images without their "highlight something in the image" thing getting in the way.
I miss being able to search for information with Google. It's just not a thing any more. Yandex seems to work but the association with Russia is off-putting given Ukraine.
Any recommendations for a replacement, especially one that considers AI generated plausible looking noise to be spam?
If you’re willing to pay, Kagi. They include AI features, but they are optional. I did no real research, but I barely get typical AI generated results when I search for something.
One isn’t even search, it’s generic questions: “how do I make pickled onions” type queries where an LLM can actually make sense of the millions of recipes and summarize them for you. I guess this is the vast majority and what google has been drifting towards for years now as the average user has become less technical.
But we need a completely separate one, based purely on indexing, for “very specific error code stack trace” type searches where LLMs are zero use and current google results are useless. Perhaps an LLM could be trained to separate the blogspam from the actual information to help keep the dataset clean from the inevitable SEO-poison. There’s at last no lack of negative examples in the data…
Search will be dead in a year or two. Nobody wants to double check information, they just want fast and easy answers, unless they're intentionally navigating to a specific website (facebook, electricity provider, etc)
I recently built a pretty simple static site. I wanted to know how bad the search engines were out to lunch. I chose an industry and domain, then built something simple using a template. Then I did what you should never do with SEO. Keyword stuffing, geo tagging and stuffing, using duplicate content and then copying material directly from high ranking sites in the same industry.
NONE of it was penalized and the site within a week or so was on the top two pages for dozens of specific searches I very deliberately abused SEO in order to achieve.
Nobody is awake right now at wheel. Not long ago, my site would've been buried in a matter of hours. Now? I'm rewarded when I purposefully manipulate the SERPS for some research and get listed on the front page.
If people think AI is going to make this better, they're in for a rude awakening. Its only going to get worse as more people realize how easy it is to manipulate the search engines to get your stuff ranked higher and take advantage of it. There doesn't seem to be any checks on the validity of what the AI is serving up - making it ripe for abuse.
That's a really interesting experiment. I've been pretty down on Google for a while now but that result is even worse than I would have expected. It's shocking how little Google cares about their core product. Same for Microsoft/Bing, but at least for them it's not a core competency.
Not only that but even for topics not abused to death by SEO, Google still manages to include absolutely irrelevant pages.
It seems like they have had some system for years already which includes an absolutely unhealthy amount of pages that in their algorithms view are similar to the thing I ask for without being the thing I ask for.
My go-to example is when I search for a construct from a JS framework that a previous dev used and Google gives me lots of results that seem superficially relevant but when I open them I think half or more of the results at the first page are about other, similar frameworks.
This is extremely annoying because they manage to hit a spot were the results are close enough that I often have to read a bit before I realize it.
Exactly. Who wants to double check information? You do it out of necessity.
If Google or its successor can give you exactly the information you need, with references/citations to appropriate sources/alternatives that you can spot check for a sense of security, then everyone is going to use it, and only it.
90+% of "AI" use by the general public is going to happen via Google and Microsoft products. It was always going to be a feature of established, scaled businesses with cloud offerings.
People that think OpenAI is going to replace, or even meaningfully effect, Google are ignoring basic laws of economic gravity.
I guess I might give it a try, haven't used Google search in a long time though. It used to be great, now just seems to show keyword matches that have the most ads.
For sure Google is not great at AI given the lower and lower quality of their results in the last months.
I hope not to be the only one to have noticed this...
Compared to Bing CoPilot, Google feels dated now. I know it sounds like a crazy thing to say, but being able to chat with your search results and refine what you are looking for. Get summarized explanations with sources is just wow.
I've tried CoPilot a few times and it sucked for me. The last attempt was when I was trying to find a text document where I documented a bunch of SMART data. I asked it if it could help me find a document with SMART data in it and of course it says "yes!" Then it proceeded to give me the dumbest suggestions I've ever seen in my life. The dumbest was to use CrystalDiskInfo to search for documents containing SMART data.
The only innovation AI is making from what I see is that it can replace people and doesn't cost anything. It's dumb as bricks and can't say "I don't know."
When I look up something as basic as time zones in two different regions, I don't want AI to imagine something, I just want the results from the thousands of sites available out there to just be presented to me.