How do ChatGPT, Perplexity and Google AI choose their sources?

How do Google AI Overviews and AI Mode choose sources?

Google's own documentation is the place to start. In AI features and your website, Google says there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimisations are necessary. A page must be indexed and eligible to be shown in Google Search with a snippet. Google's longer guide adds one condition: the site must be included in Search generative AI features in Search Console, which is the default setting. That is the technical bar.

The same page says structured data must match the visible text, that important content should be available as text, and that Business Profile and Merchant Center information should be kept up to date where they apply. You do not need to create new machine-readable files, AI text files, or special markup to appear in these features.

Google's longer guide to generative AI features (last updated 10 July 2026) adds that these features are rooted in the same core ranking and quality systems as Search. It describes two techniques in plain terms: retrieval-augmented generation, which pulls current pages from the Search index to ground a response, and query fan-out, which runs related searches to gather more supporting pages. From Google's point of view, the work of being a useful source in these features is still SEO.

None of that is a promise that a given page will be used. Google also says that meeting the requirements does not mean a page will be crawled, indexed or shown.

How does ChatGPT search choose sources?

OpenAI publishes the crawlers, not a ranking formula. On its crawler documentation:

• OAI-SearchBot is used to surface websites in ChatGPT's search features. If a site opts out, OpenAI says it will not be shown in ChatGPT search answers, though it can still appear as a navigational link. A robots.txt change can take around 24 hours to take effect.
• GPTBot is used to crawl content that may be used in training OpenAI's generative models. Allowing or blocking it does not, by itself, decide search visibility.
• ChatGPT-User may fetch a page when a person asks a question. OpenAI says this is not an automatic crawl, that robots.txt rules may not apply, and that this crawler is not used to decide whether content appears in Search.

OpenAI says ChatGPT search sometimes uses other search providers as well as these crawls. Beyond the crawler rules, OpenAI does not publish a list of signals that decide which page is cited. If a claim about a search partner or a hidden ranking factor cannot be checked against a primary source, treat it as a guess.

On 10 November 2026 OpenAI also launched Businesses in ChatGPT in beta, which lets eligible businesses claim a profile and update key details. That is a way to keep facts current. It is not a published method for being cited in an ordinary answer, and claiming a profile does not mean the business will be named.

How does Perplexity choose sources?

Perplexity documents two crawlers on its crawler page:

• PerplexityBot is used to surface and link websites in search results on Perplexity. Perplexity says it is not used to crawl content for AI foundation models, and that a robots.txt change may take up to 24 hours to be reflected.
• Perplexity-User may visit a page when a person asks a question, so the answer can be accurate and can include a link. Because the fetch is user-initiated, Perplexity says this fetcher generally ignores robots.txt rules.

In the product itself, Perplexity shows citations inline. That makes it easier to see which pages were used in a given answer. It does not publish a formula for how those pages were chosen. As with the other engines, a page that cannot be reached or read is a weak candidate.

What do the three have in common?

The documented overlap is modest, and it is practical:

• The page has to be reachable. A robots.txt rule, a firewall challenge, or text that only appears after a script runs can take it out of consideration.
• For Google, and as a sensible check for the others, the page should be indexed by the search systems the engine draws on.
• The page should state the fact in plain words, close to the question a person asked.
• The same facts should appear on the site and on the profiles the business already controls, so the engine is not left to reconcile conflicting descriptions.

That is the work described in the guides to what AEO is and what GEO is, and in the GEO service.

What do they not use, or say they do not use?

Google is the most explicit. Its guide says you can ignore, for Google Search, several ideas that circulate as “AEO” or “GEO” tactics:

• llms.txt and similar files. Google Search does not use them. Keeping one “will neither harm nor help” visibility or rankings in Google Search.
• Special schema for AI. Structured data is not required for generative AI search, and there is no special schema.org markup you need to add. It remains useful for rich results, and it must match the visible text.
• Inauthentic mentions. Seeking them “isn’t as helpful as it might seem”. Google’s quality and spam systems apply to these features too.

OpenAI’s and Perplexity’s crawler pages do not say that an llms.txt file is how they choose sources. There is no documented evidence that publishing one will get a page cited.

How we use this in the work

We use AI tools for research, content optimisation and AI-search monitoring, and we study how ChatGPT, Perplexity and Google AI choose what to cite. We do not promise a ranking, a traffic figure, or a mention in any answer engine.

In practice that means we read the public pages, ask the questions a buyer would ask, record what each engine says and which URLs it cites, and set out the fixes in writing. The method is described on how the work runs. It is a way to make the site easier to find and to read. It is not a way to control the output.

Questions

Does schema get me into AI Overviews?

No. Google says structured data is not required for AI Overviews or AI Mode, and that there is no special schema to add. It can help a page be eligible for rich results in ordinary Search. Any markup must match the visible text.

Should I block GPTBot?

That is the site owner's choice. GPTBot is used for training OpenAI's generative models, not for deciding ChatGPT search visibility. Blocking OAI-SearchBot is the change that, according to OpenAI, stops a site being shown in ChatGPT search answers. The two rules are independent. A business can allow search and still disallow training, or the reverse.

Can you make ChatGPT cite us?

No. We can make the public pages easier to find and to read, and we can show you what the public answers already say. We cannot make ChatGPT, Perplexity or Google AI cite a page.

Write to us

If you want a written check of the public pages and the public answers, write to hello@clearlyseen.co.uk. Name the business and the sites. Please do not send personal details about your customers, patients or clients.