How do you get cited by ChatGPT, Perplexity and Google's AI Overviews?

What is GEO?

GEO, short for Generative Engine Optimisation, is the work that gets a website picked up and cited in the answers of AI-based engines: ChatGPT, Perplexity, Google's AI Overviews and AI Mode. Where SEO aims for a place in a list of links, GEO aims for a mention, often with a link, inside a written answer.

The two overlap heavily. Each engine first has to be able to read your page, then judge it reliable and precise enough to cite.

Which crawlers should you let in?

This is the most concrete point, and the one most often missed. Each engine has its own crawlers: some are used to train models, others to show sites in answers. According to their official documentation:

Two useful nuances. At OpenAI, ChatGPT-User visits a page when a user asks for it, but it doesn't decide whether a site appears in results. At Perplexity, Perplexity-User does the same and generally ignores robots.txt, since a user requested the page.

A website that blocks OAI-SearchBot, often through an overly broad robots.txt rule or a firewall, won't appear in ChatGPT's search answers. It is the first thing I check.

What does Google require for its AI Overviews?

Nothing more than for standard search. On its page about AI features ↗, Google writes that there are no additional requirements: the page must be indexed and eligible to show with a snippet. It also states that you don't need to create machine-readable files, AI text files or special markup.

Google adds two things worth knowing:

• AI Overviews only show when they add something to standard search, which means they often don't appear;
• they rely on several searches run in parallel, which lets them cite a wider range of sites than the top ten results.

To limit what Google uses from your pages, the controls are the same as ever: nosnippet, data-nosnippet, max-snippet or noindex.

Is the llms.txt file useful?

llms.txt is a proposal by Jeremy Howard, published in September 2024 on llmstxt.org ↗: a Markdown file at the root of the website that summarises the business and lists the important pages for AI tools.

Let's be clear about its status. Google says it doesn't need one, and neither OpenAI nor Perplexity say they use it in the documentation I read. I publish one on this website because it costs little, it is generated automatically from the same data as the pages, and it may help the tools that do read it. But I never present it as a way to get cited.

How do you write a page an AI can cite?

Engines cite passages that answer a question on their own. This is the format I use for the articles on this blog:

1. A direct answer at the top, in 40 to 60 words, understandable without the rest of the page.
2. Headings phrased as questions, the kind someone would type into an engine.
3. Tables whenever there is a comparison.
4. Sourced figures, linked to the official text, rather than general claims.
5. A named author and a date, so the information can be dated and attributed.

None of this is specific to AI. These are the qualities of a good reference page, which Google also rewards in its standard results. The difference is that the slightest vagueness takes you out of the answer, where a list of links still left you a place. On the technical side, the basics are still SEO: an indexed page, a fast one and a well-described one.

How do you measure your presence in AI answers?

This is the least mature part:

• Google counts clicks from AI Overviews and AI Mode in the Search Console Performance report, without separating them from the rest.
• Bing launched an "AI Performance" report in Webmaster Tools in February 2026, in preview.
• ChatGPT and Perplexity offer no equivalent tool for website owners.

The most reliable method is still manual: a fixed list of questions your customers would ask, tested every month in each engine, noting who gets cited. That is what I set up in the GEO part of my SEO work.