Gemini, is this true?
Today I searched for a CD ISO filename on Google. Just the first part of the ISO name. I knew it was an ISO for Win 7, just not which one. I wanted to find a site that had the ISOs in an archival listing so I knew which version it was.
Google gave me zero results except AI telling me what I already knew. It’s a filename.
Zero.
Google is garbage. Google search is dead.
Only Google’s AI?
i don’t know why Google would be okay with their AI destroying those Promotional links.
Sunk cost fallacy
I’ve noticed an interesting phenomenon: ChatGPT’s search function is substantially better than Google’s AI search. And Google’s AI search is, frankly, so broken that it can barely be considered functional in some situations.
I think part of the problem is that Google is asking its AI to serve two fundamentally conflicting purposes. It doesn’t just want an LLM that does LLM things. It wants an LLM that also functions as a marketing engine. And that makes sense from Google’s perspective because Google is, at its core, an advertising company that happens to operate some enormously successful technology platforms. Advertising is how Google makes the overwhelming majority of its money, so naturally it wants its AI to become part of that ecosystem.
The problem is that optimizing an AI for advertising inevitably compromises its usefulness as an information system. The AI’s job should be to determine what information is most relevant to the user’s question and present it as accurately and efficiently as possible. An advertising platform has an entirely different objective: determine what can be sold, who will pay for access to the user, and how to maximize engagement and revenue. Those objectives can overlap, but they are not the same objective.
And this gets even more interesting when you consider what increasingly capable AI systems are supposed to become. The fundamental value of an LLM, AGI, or whatever comes after AGI is that it can act as an intermediary between a human being and an enormous body of information, reasoning about what the human actually needs rather than simply presenting whatever happens to be profitable to show them.
You can certainly put advertisements around an AI. You can even have an AI identify products or services that are relevant to a user. But the moment the AI itself becomes financially incentivized to favor one answer, product, company, or piece of information over another, you have fundamentally compromised the thing that makes the AI useful in the first place.
A calculator doesn’t work better because the person who made it gets paid when you enter a particular number. A search engine shouldn’t give you a different answer because someone paid it to prefer that answer. And an intelligent system certainly shouldn’t be quietly optimizing its reasoning around whoever is willing to pay for access to your attention.
The irony is that the more capable AI becomes, the less compatible that model should be with traditional advertising. Advertising depends on controlling what people see. Intelligence depends on being able to determine what is actually relevant. Those are fundamentally different incentives.
As usual, the limiting factor may not be the technology itself. It may simply be our inability to resist turning every useful invention into another machine for extracting money from one another.
My bran?!?
That’s what helps me poop!
Bold of them to assume I use Google.
Why does the author assume I’m dumber than them. I basically never use AI. It’s stupid and pulls shit from the Internet and just assumes it’s correct. I, on the other hand, have media literacy and can figure out which info is useful and likely accurate myself. It’s like asking a chronic liar a question. It’s a waste of time when we have DDG and a few brain cells.
The amount of times I’ve gone to a top ranked website by duckduckgo and read the first paragraph only to go “oh this is just more slop” is ridiculous. I have to append reddit to every search to get anything useful and even then it could be bots.
The article mentions but then bypasses how the ai summaries link out to aggregation websites which seem to be showing AI summaries themselves, which in turn link out to other sources than the original. The problem is therefore not only google’s ai summaries, but also the ease that ai enables these other sites to publish with the speed they do (fully automatic probably), and they’re probably optimized to be featured in ai summaries.
That’s not a fight they’re going to win by writing higher quality content, and it’s only going to get worse. I’m not sure what remediation there is.
It really sucks.
Don’t use google
“That’s not a fight they’re going to win by writing higher quality content” - Why not?
I haven’t touched a Google product in months. Pretty sure I’m good.
No YouTube or Android or Google Maps?
Ah shit, Android. You’re right. I’m using mostly FOSS/F-Droid apps and totally forgot about that. Good catch, I’m a liar.
Front ends for YT, so I never actually touch it. Try and use OpenStreetMaps/CoMaps as much as I can,
Well no, the title is wrong. Google itself destroyed Google Search. And the Internet does not depend on Google Search. And most of us don’t use Google’s AI.
Of course there are some real damages, but the title itself is bait click. It’s just false at face value. Do better next time.
Most people use Google’s AI. It has 2.5 billion monthly active users. 📈💸
Most people on the fediverse might not (although I have my doubts about that), but even among my fellow developers the majority use Google Chrome and Google search. Looking at my circle outside tech, it’s pretty much everyone.
Sadly, there is an astounding number of
stupidignorant, gullible people who have not been trained in media literacy.In fact, the problem is so epidemic we should have more organizations that create and proliferate media literacy content.
Such as If a news story sounds dubious, or if you are particularly invested in a story being true, search for other coverage from other news agencies, and (if the story isn’t verbatim) check the facts. The details that are consistent between separate coverage are more reliable than the rest.
I haven’t used Google as a primary search engine in a decade or more. Nothing I do depends on Gemini (I don’t even use Apple Intelligence that’s a Gemini white label).
Personally, I think what Gemini is destroying is a market-based financial model that many came to associate with the Internet, information access and lifestyle.
Most of the websites I visit pre-date Google and some have gone as far as blocking all bots via robots.txt. They survive not via predatory ad networks but by donations and volunteers.
It’s why I like Lemmy; it follows the same model.
Just a quick correction. Robots.txt doesn’t block anything, it simply politely asks the search engine to not look here or use what is contained within. A lot of the search engines ignore this and catalogue the contents of the website anyway. And pretty much all of the AI database downloaders ignore it with impunity.
I think it’s cute you think the AI sites are paying attention to robots.txt. 🤣
I am back to early web 2.0 in many cases , forums and such. Simply because they needlessly fucked something that worked perfectly before 2015 or so.
So, I disabled everything AI the I can think of, use DDG as a search engine, ad block wherever, blah blah blah, my search results SUCK. Is there a better way to fix it?
And I use Linux at home…no blows crap
Suck in what way? And what kind of stuff do you usually search?
I used to think search was kind of an universal thing but after looking at some people’s examples of what is a good or bad result I noticed that is definitely not the case.
For me a good search result is a list of sites that contain the words I typed on the search bar, sorted by how closely the words match and then perhaps subsorted by some website ranking system. For that, DDG has been quite decent.
Very curious whether you’re old enough to remember how google used to be in say, 2007.
I am and I do, but I didn’t mean to say DDG is just as good. It’s just been good enough for me.
Not OP, but I find the DDG results are SEO sites at the top for pages.
My search: how to fit a dog collar
Top 50 results from DDG:
fitting a dog collar is difficult and many people wonder how to fit a dog collar. This article will show you how to fit a dog collar.
A dog collar is difficult to fit. Perhaps you’ve wondered how to fit a dog collar.
To be fair, SEO ruined the internet and Internet searches well before AI and it’s not DDG’s fault. But I would love it if blatant SEO was filtered out of my results.
Part of it might be how Google trained everyone to use natural language queries in their searches.
Instead of “how to fit a dog collar” does
“pet advice” +”dog collar” “guidance for new dog owners”
give you better results?
The idea is to search for a cluster of terms that are likely to be on the page you want to find, but aren’t hyper-focused like SEO pages.
I use SearXNG it works pretty well for me.
Kagi is pretty decent but the sad truth is that something like duck.ai works better as a search engine than normal search engines
robots.txt has been summarily ignored for nearly 20 years, my friend. Sorry to bear the news.
Nah, Gemini is really bad.
There’s a push to use AI where I work and we have our own Gemini, openAI, and Grok models and they literally check to make people are at least “trying” to use it…
Gemini will spin “thinking” for maybe 30 seconds, then something like “checking google results” and then just spitting out the same “ai overview” you’d get from searching the same thing.
All Gemini can do is Google something and summarize the first few results, but at that point a human could just pick a trustworthy site themselves instead of going with what has the best search optimization all mashed up.
With that one (and likely a lot of others) the big search engines becoming shit from AI slop and search optimization is going to also make AI even shittier, then the top links the next AI will be shittier.
We’ll get AI rewrites of AI rewrites, repeating itself exponentially faster as more money is sunk into it and hallucinations just keep repeating.
Like, this is the natural results of monopolies. Corporations get so large they can’t help but ruin the parts that work by integrating parts no one wants to try and make them profitable.
Because one corporation can’t function and be as large as Google. It’s just too much to keep organized while constantly growing profits. It’s like how the square cube law limits the size of buildings and animals.
The larger an animal gets, the weaker it gets pound for pound, until it would reach a point where it isn’t even strong enough to breathe…
That’s where these giant tech companies are headed.
This is exactly the kind of circular cascade failure that we are going to see. Once the AI engines all start training on their own data and sourcing each others slop it just gets worse and worse.
If it wasn’t already an issue, AI companies wouldn’t be destroying old books to guarantee they weren’t training on AI slop…
The problem is if AI replaces writers, there will never be anything new to train AI on.
Humans haven’t hit a wall on innovation, were just no longer in the massive boom from early internet.
Historically the only way to cause those booms is to connect humans together to facilitate the exchange of ideas and collaboration.
We’re headed so fast to the opposite I don’t know if tech bros are too overconfident to understand they were a product of their time and not the other way around, or if they’re smart and selfish enough that they’re intentionally trying to freeze human innovation so that there won’t be another boom so no other group could clean up and replace them.
I don’t think today’s bots take any notice of robots.txt. It’s a relic from a more civilized time.
I run a few personal sites with some being available to the wider internet, and I have found that they generally do respect it. All the bots which advertise themselves as OpenAI, Anthropic, or Google do obey. However, I do occasionally see a stealth bot appear which uses a normal browser’s user agent and those just do whatever they want.
Meta absolutely does not respect robots.txt. they are scraping what I have at around 200 hits / minute.
I also notice there is some data center somewhere (or multiple) proxying all its requests through residential proxies so I can’t tell who is scraping. They don’t scrape like Meta though. Far slower. These are the stealthy bois using spoofed user agents you mention. Can’t tell who they are, though. How do they get so many residential IPs?
I have no organic traffic (I am literally just running a crawler tarpit. My page has nothing) so its really obvious that Im watching AI scrappers. My tarpit generates random links that all resolve to the same place so its super obvious its not human. Its also super obvious when two distinct IPs follow the same random word salad link milliseconds apart.
It is a business model. People install free apps on their phone. These apps generate their income by acting as residential proxy for scrapers. They form a botnet of millions of ips, actual phones, without the owners knowing it.
I’ll have to double check my stats later. Haven’t looked at it for a while.
I certainly don’t recall them ignoring robots.txt, but I also don’t remember seeing Meta in my dash at all, so it’s entirely possible they, for wherever reason, never found my site.
They don’t—but they DO use search index data to figure out where to crawl.
It’s worse than that. Not only are people going with the AI Overview as gospel, the search results themselves have nothing to do with the normal SEO that’s been the standard for the last 20+ years. People are picking the AI Summary, because the search itself is fundamentally broken.
Gone are the days that website content + traffic determine the order or relevance. Now websites have to include specifically formatted fluff that the AI summarizers like in order to show as relevant results. Just ask Louis Rossman.
The problem with search is not that it’s broken. The problem is that an effective search system like google had on the early days is only truly effective for “power users” who try to understand how to use search effectively.
The majority of people have been using Google as if it was an AI long before it turned into one. People search for questions expecting to see answers, so search engines tried to adapt to that. And then when you’re trying to guess what the user wants to search for instead of just searching literally for what they asked, your search engine will never be 100% effective for anyone.
Like, on the old days for each person who could get great search results from Google you had 9 people who failed to use Google at all. That was one happy user and 9 very unhappy ones. Nowadays they have all 10 of them getting mediocre to ok-ish results. They are never getting anyone as satisfied as that one person used to be, but in general the average satisfaction has gone up.
We’re entering a dark age that’s being marketed as a renaissance.
