Some of it is documented by the companies running these systems, and that part is worth acting on today. The rest is confident invention, including one instruction repeated everywhere that does the opposite of what it claims. This page separates the two and gives you free tools to check your own site against the half that is real.
You will read that blocking Google-Extended in robots.txt keeps you out of Google's AI. It does not. Google-Extended governs whether your content trains Gemini. AI Overviews is served from ordinary Googlebot indexing, so a site that blocks Google-Extended is still eligible to appear in one, and has meanwhile opted out of something else entirely without meaning to. If a supplier gave you that advice, it is worth asking what else on your robots.txt they were guessing about.
The half that is real
Each of these is checkable against the operator's own documentation, which is the only reason any of it is on this page.
OpenAI
GPTBot collects content that may train the models. OAI-SearchBot is the one that decides whether you can appear as a cited source inside ChatGPT's search answers. ChatGPT-User only fires when a real person pastes your link or a custom GPT calls your page. They take separate robots.txt directives, so you can allow search visibility while refusing training, and a site that blanket-blocks OpenAI has usually refused all three by accident.
Google-Extended governs whether your content trains Gemini. AI Overviews is served from ordinary Googlebot indexing, so blocking Google-Extended changes nothing about whether you appear in an AI Overview. If somebody told you to block it to stay out of Google's AI, they were wrong, and the thing that would actually remove you is blocking Googlebot, which removes you from search entirely.
Since June 2026
Search Console gained a Search generative AI setting that genuinely excludes a property from AI Overviews, AI Mode and generative features in Discover. It reached UK properties first, under pressure from the Competition and Markets Authority. It is worth knowing it exists and worth leaving alone: excluding yourself removes the impressions and the traffic that come with those surfaces, and you cannot be cited by a surface you have opted out of.
ChatGPT
ChatGPT's web results have documented reliance on Bing's index, which makes Bing Webmaster Tools worth the twenty minutes almost nobody spends on it. It is not the whole story: independent researchers have found evidence of Google index use as well, which OpenAI has not documented. Anyone telling you Bing is the single answer is simplifying something that is not simple.
Google, 15 June 2026
Google said so directly. No engine has confirmed reading one, and the people selling it as a ranking lever are guessing. We publish one, we built a free generator for it, and we still say this, because the honest case for the file has never been ranking: it is the one place your business is described in your own words rather than inferred from your marketing copy.
Structured data
Organization and LocalBusiness markup publish your name, address, hours, prices and contact as data rather than as sentences a machine has to interpret. That does not buy you a mention. It removes the guessing from the part of the answer that is about you, which is the part you would least like guessed.
What you can actually control
Before anything else. A robots.txt written in 2019, or a firewall rule nobody remembers, is the most common reason a site is invisible to assistants, and it is invisible in a way that produces no error and no warning. Our free checker reads your robots.txt and tells you which assistant crawlers are permitted, and it is honest that it cannot see CDN or firewall rules, which is where accidental blocking most often lives.
Run the AI crawler checkYour address, opening hours, telephone number and the kind of business you are, published as JSON-LD rather than left in a footer for something to infer. Our generator produces valid LocalBusiness markup in the browser, with nothing sent anywhere, and it deliberately has no aggregate rating field because typing in ratings you have not earned is the fastest way to have your structured data ignored everywhere.
Generate your schemaAn llms.txt is a short plain-text description of what you do, what it costs and which pages matter. No engine has confirmed reading one and we will not pretend otherwise. It costs about an hour, and it is the only place your description is yours rather than assembled from whatever an assistant happened to read.
Write your llms.txtA citation. There is no submission form, no placement and no paid inclusion in an assistant's recommendation, and any agency quoting you a price for one is quoting for something it does not control. You will also see confident numbers going round about how many reviews you need and what star average gets you named. Those are somebody's marketing, not measurement, and you will not find them on this page. What is genuinely in your hands is being reachable, being consistent, and being explicit about the facts you would least like guessed. That is the whole of the honest offer, and it is worth doing because most of your competitors have not done it.
The long versions are written up in the engineering notes: why GPTBot reads your sitemap first, JSON-LD that AI engines can actually use and the llms.txt field guide.
