How to Get Cited by ChatGPT, Perplexity and AI Overviews

To get cited by ChatGPT, your page has to be reachable by its search crawler, OAI-SearchBot, and good enough to rank in the results ChatGPT pulls before it writes an answer. Then the page needs short, self contained passages that answer the question with specific facts. The same three steps work for Perplexity, Claude and Google AI Overviews: allow the right crawler, rank, and write liftable answers.

If you want ideas like these working inside your own business, see our AI development and automation service.

Generative engine optimisation (GEO) is the name for that work. It is mostly classic SEO plus two jobs: keeping AI search crawlers unblocked, and writing so a model can quote you without rewriting you. Facts here are as of November 2026.

How does ChatGPT choose which sources to cite?

ChatGPT cites pages it finds through a live web search at the moment you ask, not pages it memorised in training. According to OpenAI's crawler documentation, OAI-SearchBot is the robot that surfaces websites in ChatGPT's search features, and sites that opt out of it "will not be shown in ChatGPT search answers", though they can still appear as plain navigational links.

ChatGPT also partners with outside search providers. OpenAI's help page says it rewrites your question into targeted queries, sends them to those providers, and sometimes runs a second, narrower round; it names Microsoft, which runs Bing, among them. So two things decide whether you show up: crawlability for OAI-SearchBot, and ranking in the index ChatGPT queries.

AI search is retrieval augmented generation at web scale; our plain language guide to RAG for business leaders explains how a model reads retrieved pages and writes from them.

How do Perplexity, Claude and Google AI Overviews pick sources?

Each engine runs its own search step first, then writes an answer from the top pages it retrieved. The differences are in which crawler feeds which index.

• Perplexity. Its crawler documentation says PerplexityBot is "designed to surface and link websites in search results on Perplexity" and is not used to crawl content for AI foundation models. A second agent, Perplexity-User, fetches pages when a person asks a question, and because a user triggered the fetch it "generally ignores robots.txt rules".

• Claude. Anthropic lists three bots on its crawler help page. Claude-SearchBot indexes content to improve search results, Claude-User fetches pages when a Claude user asks something, and ClaudeBot collects content that may be used for training. Anthropic says blocking Claude-SearchBot "may reduce your site's visibility and accuracy in user search results".

• Google AI Overviews and AI Mode. These use the normal Google Search index. Google's guide to AI features and your website states that to be shown as a supporting link, a page "must be indexed and eligible to be shown in Google Search with a snippet", and that "there are no additional technical requirements." Google adds that both features may fan out into several related searches across subtopics, so they often link pages that do not rank first for the original query.

The practical rule: for Google, rank well in Google. For ChatGPT, rank well in Bing and allow OAI-SearchBot. For Perplexity and Claude, allow their search bots so they can build their own copy of your page.

Which crawlers do you need to allow in robots.txt?

Allow the search crawlers and decide separately about the training crawlers. Every vendor now splits the two, so you can appear in AI answers without giving your content to model training.

1. Allow OAI-SearchBot for ChatGPT search. OpenAI notes it can take about 24 hours after a robots.txt change for its systems to adjust.

2. Allow PerplexityBot for Perplexity answers.

3. Allow Claude-SearchBot for Claude search results.

4. Allow Googlebot. Google says robots.txt rules for Googlebot are the control for AI features in Search, because AI is built into Search itself.

5. Decide on GPTBot, ClaudeBot and Google-Extended. These govern training. Blocking them does not remove you from AI search answers.

Google-Extended confuses almost everyone. Google's common crawlers documentation says it governs Gemini training and grounding in Gemini Apps and Vertex AI, and "does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search." Blocking it will not keep you out of AI Overviews. To limit what AI Overviews show from a page, Google points to nosnippet, max-snippet and noindex instead.

robots.txt is only half the gate. A CDN or firewall rule that challenges unknown bots blocks these crawlers even when robots.txt says yes, which is why OpenAI and Perplexity publish IP ranges to allowlist. The common accidental blocks, including GPTBot versus OAI-SearchBot mixups, are in you probably blocked the AI shopping agents by accident. To test your domain in a minute, run the free GEO checker.

How do you write passages that AI engines quote?

Write each section so its first two or three sentences answer one question completely, with a specific fact, and still make sense when lifted out alone. A model pulls passages, not whole pages, so a paragraph that leans on the one above it is hard to quote.

• Answer first. Put the direct answer in the first sentence under each heading. Context and caveats come after.

• Self contained paragraphs. Name the subject again instead of writing "it" or "this tool". A reader who lands on paragraph six should not need paragraph five.

• Specific numbers with a source. "Allow OAI-SearchBot; OpenAI says changes take about 24 hours" beats "update your robots.txt and wait a bit".

• Headings phrased like questions. People type questions into AI assistants, and matching headings help the engine map your passage to them.

• Text, not pictures of text. Google lists "making sure that important content is available in textual form" as a best practice. Prices and specs inside images or client side JavaScript are invisible to many crawlers; see what the DOM is and why AI agents need it.

There is research behind this. The GEO paper by Aggarwal and colleagues, mostly from Princeton University (arXiv 2311.09735, KDD 2024), tested nine ways of rewriting source pages across 10,000 queries on an engine built on gpt-3.5-turbo, then checked results on Perplexity. Adding quotations, statistics and cited sources were among the strongest methods; the best improved visibility over the baseline by 41% on one metric and 28% on another. Keyword stuffing did worse than doing nothing. The authors also found the gains were largest for lower ranked pages: citing sources raised visibility by 115.1% for pages ranked fifth in the search results.

Read those numbers with care. The engine was a 2023 research setup, not today's ChatGPT, and the paper says results vary by domain. The direction still holds: evidence, quotes and sources help; keyword padding does not.

Does structured data help you get cited by AI?

Structured data helps the engines understand what a page is, but no vendor documents it as a citation switch. Google's AI features guide asks only that your structured data "matches the visible text on the page", and it repeats that no special markup is required for AI Overviews.

Use schema for what it was always good at. Organization markup ties your brand name to your site and profiles. Product markup with price and availability feeds AI shopping answers, and Google tells merchants to keep Merchant Center data current. Article markup with an author and date signals freshness. Sellers should read agentic commerce in 2026: what the numbers say.

Do brand mentions on other sites matter for AI search?

Yes, because the engines answer from many pages at once, and the pages they retrieve about you are often ones you do not own. Ask "best invoicing tool for freelancers" and the top results are comparison articles, review sites and forum threads. If your brand is absent there, it is absent from the answer.

The work is old fashioned PR. Get listed accurately in the comparison pages that already rank for your category. Publish original numbers other writers will cite, since a statistic that starts with you carries your name into every article that repeats it. Keep your company and product names identical everywhere so engines know the mentions refer to one entity.

What is llms.txt and does it help you get cited?

llms.txt is a proposed Markdown file at /llms.txt that gives AI agents a short, curated map of a site, with links to clean Markdown versions of important pages. It was proposed by Jeremy Howard at llmstxt.org in September 2024 and is now on version 2 of the proposal.

It is a proposal, not a standard. The proposal page says thousands of sites publish one, including OpenAI, Anthropic and Google for their developer docs, and Chrome's Lighthouse now checks that a site serves one without a server error. It also says the file is used most heavily for software documentation, by coding agents. None of the crawler documentation from OpenAI, Perplexity, Anthropic or Google says their search systems use llms.txt to choose citations.

Our take: publish one for developer docs, because coding agents read them. For a marketing site it probably will not move citations, so fix crawler access and writing first.

GEO vs SEO: what actually changes?

GEO keeps the SEO foundations and changes the goal: being quoted, not only clicked. Indexing, internal links, page speed and helpful content still decide whether you enter the candidate pool. GEO adds three things.

• More crawlers to manage. You now have four or more search bots to allow and several training bots to decide on.

• Passage level writing. A page that ranks but buries its answer in paragraph nine rarely gets quoted.

• Off site presence. Third party pages carry more weight because the engines fan out across many sources per answer.

What we see when we build sites is that the first two live in a few files: robots.txt, firewall rules and page templates. That is why we set them up in our web development work from day one.

How do you measure whether AI engines cite you?

Measure in three places, because no single tool reports AI citations across engines.

1. Google Search Console. Google includes AI Overviews and AI Mode traffic in the Performance report under the Web search type. A click inside an AI Overview counts as a click, and every link in it shares one position. You cannot filter for AI Overviews alone, so watch queries where impressions rise but clicks fall.

2. Referral traffic in analytics. Filter sessions by referrer for chatgpt.com, perplexity.ai, claude.ai and gemini.google.com. Volumes are small, but the trend shows whether your changes landed.

3. A manual prompt panel. Write 20 to 30 questions your buyers really ask. Monthly, ask each in ChatGPT, Perplexity, Claude and Google, in a fresh session, and log whether you are cited and who is cited instead.

A hypothetical worked example. A small software company is never cited by ChatGPT for "best tool for X". Its panel shows two competitor comparison articles cited every time, and its server logs show no OAI-SearchBot visits because a firewall rule challenges unknown bots. Fix one: allowlist the published IP ranges. Fix two: a comparison page with a direct answer up top in plain text. Fix three: pitch an accurate entry to the two cited articles. Rerun the panel in four weeks.

Key takeaways

• AI engines cite pages they retrieve through live search, so ranking still comes first.

• Allow OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot; treat GPTBot, ClaudeBot and Google-Extended as training choices.

• Blocking Google-Extended does not remove you from AI Overviews.

• Answer each question in the first sentence under its heading, with a sourced number.

• Quotes, statistics and cited sources raised visibility in the GEO paper; keyword stuffing lowered it.

• llms.txt is a proposal with no documented effect on citations.

An automated version of that prompt panel is the kind of thing our AI agents and automation team builds.

Frequently asked questions

How long does it take to start appearing in ChatGPT answers?

OpenAI says robots.txt changes for OAI-SearchBot take about 24 hours to register. After that, appearing in answers depends on ranking in the search results ChatGPT queries, which can take weeks for a new page. Expect a month before a prompt panel shows a reliable change.

Does blocking GPTBot stop ChatGPT from citing my site?

No. GPTBot controls whether your content may be used to train OpenAI's models. OAI-SearchBot controls whether you appear in ChatGPT search answers, and OpenAI treats the two settings as independent.

Can I opt out of Google AI Overviews but stay in normal search?

Not with a robots.txt token. Google says Google-Extended does not affect Search, and AI features are part of Search. You can limit what is shown with nosnippet, data-nosnippet or max-snippet, but those also shorten your regular search snippets.

Do I need an llms.txt file to show up in Perplexity?

No. Perplexity's crawler documentation asks you to allow PerplexityBot in robots.txt and permit its published IP ranges. It does not mention llms.txt.