Google made it harder for external models to access the web's depth.

KitaYama

Well-known member
Local time
Today, 12:18
Joined
Jan 6, 2022
Messages
2,209
This may be of interest for the site owner.

You can no longer view 100 search results at once. The default max is now just 10.
Google has quietly changed how search results appear, reducing the maximum results per page from 100 to just 10, a move that could reshape SEO, AI training, and digital visibility.
This shift means 88% of websites have already seen a major drop in impressions. By cutting access to results ranked 11–100, Google just limited what both users and AI models can “see.” Platforms like Reddit, which often ranked in that range, were hit hard, with noticeable declines in visibility and engagement.

The impact was immediate.
According to Search Engine Land, 88% of sites saw a drop in impressions.
Reddit, which often ranks in positions 11-100, saw its LLM citations plummet. Its stock dropped 15%.
This isn't just SEO fallout. It's an Al supply chain issue.

Most large language models such as OpenAl, Anthropic, and Perplexity rely (directly or indirectly) on Google's indexed results along with their own crawlers.
By cutting off the "long tail" of results, Google just reduced what they can see by 90%.

The internet's "training data" just got shallower.

For AI startups and developers, this change goes far beyond SEO. It affects how data is gathered, how algorithms learn, and who gets discovered online. The open web just got narrower, and those relying solely on search now face a tougher battle for attention.

 
First, thank you, @KitaYama - for a very timely post of a major change that will certainly affect a lot of things that we do on a regular basis.

Second, I wonder - and this is ONLY a guess - whether Google initiated this change after some of the successful lawsuits on how the AI companies used "pirated" sources for training their LLMs (and whatever else they use). Many suits have been filed complaining that the AI companies exceed the boundaries of the "fair use" exception for reproduction or access to copyrighted content.

Third, another speculation... I wonder if Google thinks that returning too many items in a search leads to their AIs accepting irrelevant content and thus providing responses that are commonly called "hallucinations." We have to remember that LLMs return things that are deemed statistically likely to relate to to a given question. If you bring in too many bad answers and don't challenge them, the AI will not know to discard whatever linked the bad answer to the question.
 
First, thank you, @KitaYama - for a very timely post of a major change that will certainly affect a lot of things that we do on a regular basis.

Second, I wonder - and this is ONLY a guess - whether Google initiated this change after some of the successful lawsuits on how the AI companies used "pirated" sources for training their LLMs (and whatever else they use). Many suits have been filed complaining that the AI companies exceed the boundaries of the "fair use" exception for reproduction or access to copyrighted content.

Third, another speculation... I wonder if Google thinks that returning too many items in a search leads to their AIs accepting irrelevant content and thus providing responses that are commonly called "hallucinations." We have to remember that LLMs return things that are deemed statistically likely to relate to to a given question. If you bring in too many bad answers and don't challenge them, the AI will not know to discard whatever linked the bad answer to the question.
I think, these two comments from Instagram tells the whole story:
  • In other words, google made it almost impossible for smaller sites to be seen without paying for ads.
  • This gives Google an edge over all other AI competitor companies…It’s a clever commercial move by them, as they ringfence the data for their own AI
 
If you google something, you can still see many pages of results, so Google shows your site regardless of this change.

What I've seen reported is different. You see, SEO tools track how a site ranks in search results from Google. These tools were designed using a special parameter in the URL to fetch 100 results, since Google got rid of that parameter these tools can now see only 10 results, making the tools inaccurate. So a site owner who usually knew he was ranking in, say, position 11 or more, now sees that he's not even on the list. The tools will have to be re-engineered to account for this modification to let their user base have the stats they're more familiar with. If that means they will have to fetch pages 1 through 10, then that's probably cost some extra money and it will impact its users as a result.
 
These tools were designed using a special parameter in the URL to fetch 100 results,
That was the exact thing I was saying (the url at bottom of my post). By eliminating that parameter, now AI can only see the first 10 result. So an AI, although has its own crawlers, is much dependent on Google indexing. When an AI can only access the first 10 results of Google index, it has more limited knowledge to reply, hence less links to offer.
If you are a new site owner and want AI see your pages, you have to pay to make Google show your site in the first 10 results. Or wait a long time until one of AI's crawler finds your site. That was the main reason that reddit stocks dropped in one day.
If you want to put it in a simpler word, from now on , AI sees only 1/10 of Google index and you know how many people rely on AIs and the effect it causes in the result of their queries.
 
I stopped using Google to search years ago when I realized that I was only seeing results from paid sponsors. I use DuckDuckGo.
 
AI sees only 1/10 of Google index
Out of curiosity: how is this affecting you in particular? Are you using the "search for me" tool from your AI tool a lot lately or what is going on exactly?

Note: I suggest you people don't let an AI search for you unless you don't really care too much, you don't know what sources it will cite and it may very well affect your understanding of something. TLDR: always watch where you're getting your facts.
 
Out of curiosity: how is this affecting you in particular? Are you using the "search for me" tool from your AI tool a lot lately or what is going on exactly?

Note: I suggest you people don't let an AI search for you unless you don't really care too much, you don't know what sources it will cite and it may very well affect your understanding of something. TLDR: always watch where you're getting your facts.
AI gains its information mostly from what exists on web.
Only few AI are not trained by crawling info on the web.

When you ask AI something that it's not trained for, and has not enough info about, it starts to search web (using its own index and Google index). And gives you the answer.
Go to chatGPT and ask a rare question. For example : "Why Charlie Adelson was in the airport a while back?"
You see a blinking dot, and then a message in gray : Searching web
It tells you that ChatGPT uses web search to answer. No matter I ask for it or not.

When Google limits its index to AIs, those searches are limited only to the first page of Google Index. Or they have to requery 10 times, which takes more time to response.

ChatGPT says it doesn't use Google index, but relies on Bing index. Hence it's not affected. But I've seen articles that say even Bing relies on Google.

Here's a part of ChatGPT answer:

When you see the message Searching web:​


That means I’m temporarily connecting to a web search tool — but it’s not Google Search itself.

Here’s what happens step-by-step:
  1. When you ask something that needs up-to-date or uncommon information, I call a function called web.search().
  2. That tool connects to one or more licensed web search APIs — usually Bing Search, Brave Search, Tavily, or Perplexity’s backend.
  3. The results come back in structured form (titles, URLs, snippets).
  4. I read, interpret, and summarize them for you.

So yes — I do perform live searches, but not through Google.com directly.
 
Last edited:
I asked Google: How Google's 10 result per page effects AI?

Here's the answer:

Google's shift to displaying only 10 results per page, combined with the prominent AI Overview feature, significantly impacts AI models by limiting their access to training data and altering the information ecosystem. This makes it harder for external AI systems to access the depth of the web, potentially leading to shallower and less-diverse knowledge.
Impacts on external AI models
  • Thinned training data pipeline: Many large language models (LLMs) from companies like OpenAI and Anthropic rely on Google's indexed results to feed their own retrieval systems and crawlers. By limiting results to 10 per page and removing the "num=100" parameter, Google has cut off the "long tail" of results, effectively reducing the available training data by up to 90%.
  • Reduced visibility for deeper content: Websites that typically rank outside the top 10 have seen a sharp drop in impressions and citations. For example, platforms like Reddit, known for ranking deep in search results, have experienced a decrease in citations by external AI models, which limits the diversity of information available to these systems.
  • Increased difficulty for startups: Smaller AI companies and startups are disproportionately affected by this change. Without easy access to a broad range of data, their models are forced to train on a much smaller fraction of the web, making it harder to build robust and comprehensive AI systems.
  • Creation of an "AI supply chain issue": The change transforms Google's search index from an open resource into a more closed ecosystem, which creates a data supply chain problem for other AI developers.
Impacts on Google's own AI models
  • Reinforcement of top-ranked sources: Google's own AI Overview feature, which appears above the traditional blue links, relies heavily on the top 10 search results to generate its summaries. An SEOClarity study found that 77% of AI Overviews referenced links exclusively from the top 10 web results, reinforcing Google's confidence in its established ranking system.
  • Shift from search engine to "answer engine": The combination of fewer results and the prominent AI Overview means users are less likely to click on links. This shifts Google from being an index of third-party information to a publisher of its own AI-generated content, with sources cited but rarely clicked.
  • Emphasis on quality and authority: Google's AI-driven results prioritize content from sources it deems trustworthy and authoritative. Websites that build trust signals, such as clear author profiles and transparent sourcing, are more likely to be featured in an AI Overview.
Wider implications
  • Altered SEO strategies: The traditional goal of ranking high in organic search has changed. Now, businesses must also focus on getting their content cited within AI Overviews, which requires creating authoritative, concise, and easily summarizable content.
  • Increased user engagement for specific queries: While AI Overviews reduce clicks to individual websites, Google has observed that these features can increase overall search query volume, especially for more complex or conversational questions. The company reports seeing up to a 10% increase in certain query types where AI Overviews are shown.
 
Last edited:
always watch where you're getting your facts
When I ask a question from Chatgpt, it tells me where the data have been harvested from. Obviously I can click the link and read the whole article if I like.
So the source of info is clear and it's on me to believe it or not.

1.jpg
 
That's the kind of dangers I'm warning about in my post:
you don't know what sources it will cite and it may very well affect your understanding of something

Don't get your news from LLMs. I will tell you why:

Imagine a powerful entity, maybe a politician or maybe a very rich person. Let's say you saw that entity do something very wrong, but since they have power/money, they pay the most influential mediums to create a narrative where they're not to blame. Smaller mediums read the narrative and replicate it because they have no reporters or integrity (that's the current world, btw).

Now, let's say you saw what the powerful/rich guy did. But you are a mere mortal, so, since you know no better, you go ahead and ask the LLM: "hey, why did this guy do this?"
The LLMs read these mediums and outputs "Nah, this guy did nothing wrong, here's my sources".
Fast forward a few months, the guy is your president and enslaves your people and bankrupts your economy while he gets richer and more powerful.
Yup. Can't see how this is not already going on in countless places. LOL you don't even need AI, this was already happening way before it, AI just makes it worse because people are trusting its slop blindly.
 
Last edited:
Imagine a powerful entity, maybe a politician or maybe a very rich person.....Fast forward a few months, the guy is your president and enslaves your people and bankrupts your economy while he gets richer and more powerful.
I'm sure you'll hate me, but I will open it up to you.
I'm not interested in politic. I don't listen/watch news. I've never voted in my life. I don't even know who was/is our prime minister.
I believe every politician, left or right is corrupt and I've decided not to try to change anything. Because we can't.
Even in AWF (and all others forums I'm in), I have add politic, water cooler and other branches to my ignore list.
I'm heavily addicted to social media, but their algorithms know enough about what I like and what I don't, so the only posts they show is about technology/Science/fun/pets/sports and my profession. I watch so many youtubes, but my home screen is filled with posts about car/train technology, software about my profession, tricks about CAD systems etc.

I'm sitting here (even today, a national holiday day) in front of my PC and working 13 to 15 hours a day. I really don't want to waste the remaining time stressing myself with politics.

I know your advice is accurate and I don't deny what you explained. I just want you to know that it doesn't concern me.
because biased answers may exist in politic/social cases, but 2+2 is always 4 and gravity is 9.8. It doesn't matter you ask FOX or CNA or ChatGPT or Google. They all give you the same answer if the question is not political/social one. No bias in technology and science.
I'm sure other normal people will benefit from what you kindly explained.

I was just wondering if #8 & #9 answered your "how is this affecting you in particular?" query.
 
Last edited:

Users who are viewing this thread

Back
Top Bottom