Being cited by AI answer engines: what is documented and what is not

Google's Search Central documentation on AI Overviews and AI Mode is direct. There are no additional requirements to appear in either feature and no special optimizations are necessary. To be eligible as a supporting link, a page must be indexed and eligible to be shown in Google Search with a snippet. Google's optimization guide adds that a site must be included in Search generative AI features in Search Console to be eligible.

Google also describes the mechanism in general terms. The features use retrieval-augmented generation, which relies on the core Search ranking systems to retrieve pages from the Search index, and "query fan-out", in which the model issues several related searches across subtopics before composing a response. AI Overviews appear only when Google's systems judge them additive to classic results, and Google says they often do not trigger. AI Mode and AI Overviews may use different models, so the links they show for the same question can differ.

In practice, visibility in Google's AI features rests on the same foundations as ordinary search visibility. A page that cannot be crawled, cannot be indexed or is barred from showing a snippet is not eligible.

OpenAI documents separate crawlers for separate purposes. OAI-SearchBot surfaces websites in ChatGPT's search features, and sites that block it will not be shown in ChatGPT search answers, though they can still appear as navigational links. GPTBot crawls content that may be used to train OpenAI's foundation models. The two settings are independent, so a site can allow OAI-SearchBot and disallow GPTBot. ChatGPT-User fetches pages when a user asks ChatGPT a question, and OpenAI notes that robots.txt rules may not apply to those user-initiated requests. OpenAI says a robots.txt change can take about 24 hours to affect search results.

Perplexity's documentation follows a similar pattern. PerplexityBot surfaces and links websites in Perplexity search results and, according to Perplexity, is not used to crawl content for AI foundation models. Perplexity-User fetches pages in response to user questions and generally ignores robots.txt. Perplexity also advises site owners behind a web application firewall to allowlist its published IP ranges.

One failure mode sits below the content team. A firewall or CDN rule written to stop scrapers can block OAI-SearchBot or PerplexityBot without anyone deciding to. Google's own checklist for AI features includes making sure crawling is allowed "by any CDN or hosting infrastructure", as well as in robots.txt.

No provider publishes the criteria its system uses to pick one source over another for a given answer. Advice that claims to know those weights is inference. Google's optimization guide says that no third-party tool has access to Google's internal ranking or AI systems, and tells site owners to be wary of tools that claim to use internal Google metrics.

The guide also addresses llms.txt. llms.txt is a proposal, published at llmstxt.org, for a markdown file that gives language models a curated overview of a site with links to clean versions of its pages. Google says Google Search does not use llms.txt or similar AI text files. Publishing one will neither help nor harm visibility or rankings in Google Search. OpenAI's and Perplexity's crawler documentation does not mention it. A site may still publish one for coding agents or other tools that read it, but there is no published basis for treating it as a ranking factor for any of the three answer engines discussed here.

Google lists other popular tactics it says site owners can ignore for its AI features. These include splitting content into small "chunks", rewriting pages specifically for AI systems, chasing inauthentic mentions across the web and adding special schema.org markup for AI.

An answer engine that cites an organization has to work out which organization it is. That gets harder when the trading name differs from the legal name, the registered number appears nowhere on the site, and profiles on directories, social platforms and marketplaces describe the business in different terms.

Google's Organization structured data documentation describes markup that helps Google understand an organization's administrative details and tell it apart from others. The available properties include legalName, sameAs links to profiles on other sites, vatID, taxID and iso6523Code, which can carry a Legal Entity Identifier with the prefix 0199. None of the properties is required. Google's AI features guidance adds two qualifications. Structured data must match the visible text on the page, and structured data is not required for generative AI search. Its value lies in removing ambiguity, and it helps with eligibility for rich results.

The work is unglamorous. Put the legal name, registered number and registered address on the website in plain text. Mark them up consistently, and point sameAs at the profiles you control. Keep Google Business Profile and, for retailers, Merchant Center information current. Google says both can help products and services appear in its AI responses.

Decide crawler policy deliberately. Training and search access can be set separately for OpenAI through GPTBot and OAI-SearchBot. For Google, the Google-Extended token governs use in Gemini model training and grounding, and Google states it does not affect inclusion or ranking in Google Search. Record the decision and the reason for each user agent.

Test that the policy holds at the edge. Review firewall and CDN logs for blocked requests from the published user agents and IP ranges, and check robots.txt with each provider's documented tokens.

Use snippet controls only where they serve a purpose. Google treats nosnippet, data-nosnippet, max-snippet and noindex as the controls that limit what appears from a page in Search, including its AI features. A page that blocks snippets also removes itself from eligibility as a supporting link.

Measure what is measurable. Google reports AI Overviews and AI Mode traffic within the Search Console Performance report and offers a Generative AI performance report. Track citations in other engines by sampling a fixed set of questions on a fixed schedule, and treat the results as observations rather than a ranking.

Give the engines something worth citing. Google's guide asks for "non-commodity" content based on first-hand expertise. For an established organization that usually means publishing the specific evidence, methods and data it already holds, with authorship and dates on each page.