ChatGPT-User: definition of OpenAI's user-triggered fetcher

ChatGPT-User is one of three agents OpenAI documents on its crawler page, as of October 2026. Each has a separate job:

• GPTBot crawls content that may be used to train OpenAI's foundation models.
• OAI-SearchBot crawls pages so they can be shown in ChatGPT's search features.
• ChatGPT-User fetches a page when a user's action in ChatGPT calls for it.

It is a user-triggered fetcher: it runs at answer time, for one person's request, rather than building an index ahead of time. The name you see in your server logs is its user agent string.

Say a user pastes your pricing page URL into ChatGPT and asks for a summary. ChatGPT sends ChatGPT-User to fetch that page, reads what comes back, and answers. Nothing is added to a search index because of that visit.

Three mechanics matter:

• robots.txt may not apply. OpenAI says that because these actions are initiated by a user, robots.txt rules may not apply. Treat a Disallow line as a request, not a lock.
• Your network still decides. If your firewall or CDN returns a 403 or a challenge page, the fetch fails and ChatGPT has nothing to read.
• JavaScript may not run. Vercel's crawler study found ChatGPT's crawlers download JavaScript files but don't execute them. Content that only appears after hydration may be missing from what ChatGPT-User sees.

For a walkthrough of a single request, see what happens when someone asks ChatGPT to read your page.

ChatGPT-User does not get you into ChatGPT search. It controls something narrower: whether ChatGPT can read your page when a real person points it there. That person is often a buyer comparing vendors.

If the fetch fails, or returns an empty JavaScript shell, ChatGPT may answer from what it already knows or from other sites. You lose the chance to have your own page describe your product.

The next move is a check, not an optimization:

1. Confirm your firewall and CDN let ChatGPT-User through, not only OAI-SearchBot.
2. Confirm your key facts (prices, plans, product names) are in the raw HTML before JavaScript runs.
3. Check OAI-SearchBot separately if you want eligibility for ChatGPT search.

The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away. For the full OpenAI setup, read OpenAI crawlers, robots.txt and firewalls.

• "Blocking ChatGPT-User in robots.txt keeps ChatGPT off my site." OpenAI says robots.txt may not apply to it. A firewall rule blocks it more reliably, and also blocks every user who asks ChatGPT about that page.
• "Allowing ChatGPT-User gets me into ChatGPT search." OpenAI says it is not used to determine search inclusion. That is OAI-SearchBot's role.
• "Blocking GPTBot also blocks ChatGPT-User." OpenAI states each setting is independent. A GPTBot rule governs training use only.
• "A ChatGPT-User hit means I'm being cited." A fetch shows someone asked ChatGPT to read the page. It says nothing about how the answer used it.

• Perplexity-User: Perplexity's user-triggered fetcher, which Perplexity says generally ignores robots.txt rules.
• Index-time crawlers vs answer-time fetchers: why OAI-SearchBot and ChatGPT-User need different controls.
• SEO for AI agents: the guide to user-triggered fetchers in depth.

How do I block ChatGPT-User? robots.txt is the weak option, because OpenAI says its rules may not apply to user-initiated fetches. A firewall or CDN rule that matches the user agent and OpenAI's published IP ranges will stop it. Before you do that, decide whether you want it gone: blocking ChatGPT-User means ChatGPT cannot read your page even when a prospect asks it to.

Is ChatGPT-User the same as OAI-SearchBot? No. OAI-SearchBot crawls pages so they can appear in ChatGPT's search features, and OpenAI says sites opted out of it will not be shown in ChatGPT search answers. ChatGPT-User only fetches pages in response to a user's action and is not used to decide search inclusion. You need to check access for each one separately.

Can someone fake the ChatGPT-User user agent? Yes. Any client can send any user agent string, so a log line alone does not prove the request came from OpenAI. OpenAI publishes IP ranges for its agents, but it also warns against relying on IP allowlisting alone. Verify a hit by combining the user agent with the source IP before you trust or block it.

Does ChatGPT-User see content loaded by JavaScript? Probably not. Vercel's study of AI crawler traffic found ChatGPT's crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. If your prices or product details only appear after scripts run, put them in the server-rendered HTML so a user-triggered fetch can read them.