Do AI Search Engines Rely on Google Rankings, or Use Separate Ranking Systems?

Which search systems does each AI engine draw on?

The table shows the official picture. Press reports in 2025 said OpenAI and Perplexity also used scraped Google results for some answers, which neither has confirmed as a standing arrangement.

When OpenAI launched ChatGPT search, it said the feature uses third-party search providers as well as content provided directly by partners. Anthropic lists Brave Search as a web search subprocessor in its subprocessor list. These arrangements can change, so treat the table as a snapshot, and check each company's documentation when it matters.

The practical consequence is simple: a page missing from Bing's index will almost never appear in Copilot's search answers, whatever it ranks in Google. A page missing from Brave loses Claude's main search route, and a blocked crawler cuts off that engine's own fetching.

What happens between retrieval and the final answer?

Rankings only decide which pages an AI engine considers. The answer is built in several further steps, and each one is the engine's own:

1. Query rewriting and fan-out. A long question is split into several searches. Google describes this for AI Overviews and AI Mode as issuing multiple related searches across subtopics.
2. Candidate retrieval. Each search returns a set of pages, often more than the ten a person would see.
3. Passage selection. The engine looks for the specific passages that answer each part of the question, not the page that ranks best overall.
4. Synthesis. The model writes the answer, choosing which sources to cite, often favouring a mix of sources over several from one site.
5. Personal and conversational context. Earlier turns, location and sometimes memory shape the searches and the answer.

That is why a page ranking fifth can be cited while the first result is not: its passage answered a sub-question more directly. How engines combine several retrieval systems is covered in why AI search is made up of multiple retrieval systems rather than one engine.

How closely do AI citations match Google rankings?

More closely for Google's own features than for anyone else, and less closely than it used to be. In 2025, Ahrefs found that about 76% of pages cited in AI Overviews ranked in Google's top 10 for the query. Its 2026 update, covering 863,000 keyword results after AI Overviews moved to Gemini 3, found only about 38% did. The rest split roughly evenly between positions 11–100 and pages outside the top 100. A separate Ahrefs study of 15,000 prompts found the reverse picture for other assistants. On average, just 12% of the URLs that ChatGPT, Gemini and Copilot cited ranked on Google's first page for the same prompt.

Two readings follow. For AI Overviews, ranking well still helps, but most citations now come from pages outside the top 10, probably because fan-out searches retrieve pages that rank for narrower related queries. For other assistants, Google rankings are a weak predictor, because their searches, indexes and phrasings differ. Gemini sits in the second group in that study, which reflects how differently a conversational prompt is searched compared with a short query.

Why can a page with weak Google rankings still be cited?

• It matches a sub-question. Fan-out searches are narrower than the original query, and a specific page can rank for one of them.
• It ranks in another index. A page can do better in Bing or Brave than in Google.
• Its passage is the clearest answer, even if the page as a whole is not the strongest.
• It is a source the engine favours for the topic, such as a review site, forum or reference page.
• The question is long and specific, and few pages address it at all.

Do drops in Google rankings lead to drops in AI visibility?

For Google's AI features, usually yes. They draw on the same ranking systems, so a page that falls from the top 10 to the third page becomes less likely to be cited in AI Overviews and AI Mode.

For other engines, a Google drop does not directly cause an AI drop, since they use different indexes. But the reasons behind the drop often affect them too:

To diagnose, line up the dates: Google Search Console performance and its Generative AI report, Bing Webmaster Tools, and your AI visibility tracking for the same questions. If AI visibility fell only in Google's features, the cause is probably Google-specific. If it fell everywhere at once, look for technical or content causes.

How should you divide effort between Google SEO and other engines?

1. Keep Google SEO sound. It feeds Google's AI features directly and shapes much of what other engines find on the web.
2. Check Bing indexing in Bing Webmaster Tools, and submit changes through IndexNow if your platform supports it. Bing feeds Copilot and, in part, other assistants.
3. Allow the AI crawlers you want in robots.txt and your CDN, so engines with their own crawlers can fetch your pages.
4. Answer sub-questions directly, so your pages match the narrower fan-out searches.
5. Build third-party presence on the reviews, lists and publications engines cite for your category.
6. Track each engine separately, since rankings in one index do not predict citations in another.

Checking rankings in Bing and Brave alongside Google is covered in whether traditional SEO trackers can check rankings in Bing and Brave.

What do people ask about AI search and Google rankings?

Is AI visibility just SEO under another name?

Largely for Google's AI features, much less for other assistants. SEO gets you into the candidate pool; clear answers, crawler access and third-party presence decide the rest.

Do AI engines use PageRank?

Google's systems still use links among many signals, and its AI features inherit them. Other engines have their own ranking signals, and none publish them in detail.

Should I optimize for Bing to appear in ChatGPT?

Checking that your pages are indexed and competitive in Bing is sensible, since ChatGPT uses third-party search providers. Make sure OAI-SearchBot can crawl your site as well.

Can I rank in AI answers without ranking in Google?

In other assistants, yes, and increasingly in AI Overviews, where most cited pages are no longer in the top 10 for the query itself. Ranking for the narrower related searches matters as much as ranking for the main one.