How to get cited by AI: ChatGPT, Google AI Overviews, Perplexity and more.

How to get cited by AI comes down to four conditions: the engine’s search crawler can reach your page, the page is in the index that engine searches and may be quoted, a passage on it answers one of the searches the engine runs, and other sites say the same about you. Each engine checks these its own way, so a page can be cited by ChatGPT and missing from Perplexity. That is why AI search visibility is checked engine by engine.

Cited, mentioned and recommended are not the same

An AI answer can include your company in three ways:

• Cited: the answer links to your page. This happens only when the engine searched the web, and it moves fastest, because it depends on your own pages and settings.
• Mentioned: the answer names you without a link, from training data you cannot edit or from what other sites publish about you.
• Recommended: the answer puts you on a shortlist, which leans on reviews, comparisons and articles written by others.

So a citation needs a search. OpenAI says ChatGPT “may search the web automatically when your question would benefit from current information”, Anthropic says Claude searches on “topics that benefit from current information”, and Perplexity searches for every answer. An answer written from training data alone has no sources to show.

How seven engines choose what to cite

Each engine searches its own index or borrows one, sends its own crawler, and gives you a different way to drop out.

Google AI Overviews, AI Mode and Gemini

All three draw on Google’s Search index. Google’s page on AI features and your website says a page needs no special optimization for AI Overviews or AI Mode: it must be “indexed and eligible to be shown in Google Search with a snippet.” Both may use query fan-out, “issuing multiple related searches across subtopics and data sources”, and they “may use different models and techniques”, so they can cite different pages for the same question. The Gemini app grounds its answers in the same index, and the Google-Extended token controls whether your content is used for that; Google says it “does not impact a site’s inclusion in Google Search”.

ChatGPT and Microsoft Copilot

OpenAI’s crawler documentation says OAI-SearchBot “is used to surface websites in search results in ChatGPT’s search features”, and sites that block it “will not be shown in ChatGPT search answers”. GPTBot, its training crawler, is a separate setting. ChatGPT search also “sometimes partners with other search providers”, and OpenAI’s help page points to Microsoft’s and Shopify’s privacy policies for how they handle the queries. Copilot runs on Bing: Microsoft’s guidelines say Bing and Copilot rely on “the same core crawling, indexing, and ranking foundation”, and Microsoft 365 Copilot Chat grounds its answers by sending Bing a short search built from the prompt. Check that your pages are indexed in Bing as well as Google.

Perplexity, Claude and Meta AI

These three send their own search crawlers. Perplexity says PerplexityBot “is designed to surface and link websites in search results on Perplexity” and is not used for training. Anthropic says blocking Claude-SearchBot “may reduce your site’s visibility and accuracy in user search results.” Meta says allowing Meta-WebIndexer “helps us cite and link to your content in Meta AI’s responses.” Claude and Meta also run crawlers that collect training data, ClaudeBot and Meta-ExternalAgent, and all three have a fetcher that visits a page when a user asks; Perplexity and Meta say theirs can ignore robots.txt.

Check the settings that keep you out

Before rewriting anything, rule out the settings that stop engines reading or quoting your pages. Each one applies to the whole site.

• robots.txt groups. A crawler obeys only the group that names it; Google’s documentation says “only one group is valid for a particular crawler.” So Disallow: / under User-agent: * blocks every crawler you have not named, and a named group replaces the wildcard rules rather than adding to them.
• Firewall and bot protection. robots.txt can say yes while the CDN says no. OpenAI recommends allowing its published IP ranges, and Perplexity says a firewall may need to explicitly allow its bots. A crawler that never appears in a month of server logs is blocked or has not found you.
• Search Console’s Search generative AI setting. Since 31 August 2026 every site has a control, under Settings, that includes or excludes it from AI Overviews, AI Mode and the AI features in Discover. Include is the default, but a property inherits its parent’s choice, so an exclusion on the domain property covers every section beneath it.
• Snippet and archive tags. Google says nosnippet “will also prevent the content from being used as a direct input for AI Overviews and AI Mode”, and max-snippet limits it. Microsoft says NOARCHIVE keeps content out of Copilot responses. Check the page source and the X-Robots-Tag header.
• Text a crawler cannot see. Most AI crawlers do not run JavaScript, so the answer to a page’s main question has to be in the HTML it serves.

Be the clearest answer to the searches it runs

Engines split a question into several searches. Google calls this query fan-out, and its own example turns a question about a lawn full of weeds into searches such as “best herbicides for lawns”. A buyer asking which payroll software suits a 200-person company might set off searches about price per employee and switching providers mid-year. A page that plainly answers one of those is a candidate even if it ranks for none of the buyer’s words. Then make each passage easy to lift:

• Answer in the first sentence of each section, so the passage still makes sense quoted alone.
• Say something others have not. Google asks for “non-commodity content”, not a summary that “simply restates information already available elsewhere.”
• Use the words the engine searched. The AI Performance report in Bing Webmaster Tools lists “the key phrases the AI used when retrieving content” it cited. Put them in your headings where they fit.

Do not write a page for every variation: Google says doing that “primarily to manipulate rankings or generative AI responses” breaks its scaled content abuse policy.

Get named on the sites AI answers already cite

A company’s own pages are a small share of what AI answers cite. In a University of Toronto study published in September 2025, researchers sorted the domains AI engines cited into the brand’s own site, earned media such as reviews and independent publications, and social platforms. Asked ranking questions about well-known and niche brands, the engines cited this mix:

The researchers used each engine’s developer API with web search on, and consumer questions such as “What is the best-known camera brand?”, so treat the table as a direction; the Gemini rows are not AI Overviews. In software, the closest category to B2B, Google’s US results were 43.7% brand-owned, while ChatGPT’s citations were 72.7% earned. The authors describe AI search “preferring neutral third-party evaluations”.

• Find the sources already cited. Ask your buyers’ questions in each engine and list the domains each answer cites. The same review sites and publications keep coming back.
• Get onto them, accurately: a complete listing with real reviews, a correct directory profile, an article or a quote in the trade publication that keeps appearing.
• Weigh the engines differently. Perplexity drew far more on social platforms than ChatGPT, which barely cited them, so forums and video channels count for more there.
• Keep it genuine. Google says “seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem.”

How to see whether AI cites you

• Google: Search Console’s generative AI performance report shows how often your URLs appeared in AI Overviews and AI Mode, by page, country, device and date.
• Microsoft: the AI Performance report in Bing Webmaster Tools counts your citations in Copilot and Bing’s AI summaries, page by page.
• The other engines: none publishes a citation report, so count the visits. Google Analytics’ help shows how to build an “AI assistants” channel for referrals from ChatGPT, Perplexity, Gemini, Copilot and Claude, placed above Referrals.
• Every engine: a fixed set of buyer questions, asked the same way every month, with the cited domains logged.

Check each engine on its own. In the Toronto study, agreement between pairs of engines on the same questions ranged from 71% to 81%. A page cited in one engine and missing from another can point to that engine’s row in the first table rather than to the writing.

Mistakes that cost citations

• Blocking search to stop training. Block GPTBot or ClaudeBot to keep content out of training, and leave OAI-SearchBot, Claude-SearchBot and PerplexityBot open, because search and training crawlers are separate settings. Google-Extended is the exception: it covers Gemini training and Gemini’s answers together.
• Adding files and markup for the engines. Google says there is “no special schema.org structured data that you need to add”, and Google Search ignores llms.txt.
• Letting pages go stale. Microsoft says regular updates help AI systems reference the current version of a page. Show the date a page was updated, and use IndexNow to tell Bing when it changes.