Definition
ChatGPT search is web search inside ChatGPT, answering from current pages with links to them rather than from training alone. OpenAI launched it on 31 October 2024, opened it to all logged-in users that December, and removed the signup requirement on 5 February 2025.
Example. Ask ChatGPT for this week's opening hours and it fetches a page and cites it. Ask it something timeless and it may answer from memory with no link at all.
You can check whether the crawler behind it is allowed on your site with the free AI Crawler Access Checker.
Which crawler fetches pages for ChatGPT search?
The crawler is OAI-SearchBot, which OpenAI documents as the bot used to surface websites in ChatGPT's search features. It is a different bot from GPTBot, which collects training data, and from ChatGPT-User, which fetches a page because someone asked for it.
That split is the practical point, and it is the one most write-ups miss. Blocking GPTBot does not remove you from ChatGPT search, and blocking OAI-SearchBot does not opt you out of training. They are separate decisions, made in separate robots.txt groups. Our AI crawlers page lists all three with what each one does.
How do you appear in ChatGPT search?
You appear by being fetchable by OAI-SearchBot and worth quoting when it arrives. Everything downstream depends on access first: a blocked page cannot be cited, however good it is. After that it is the ordinary work of answering the question plainly, close to the top of the page.
What do people get wrong about ChatGPT search?
- Treating it as one thing with ChatGPT itself. An answer with citations used search. An answer without them came from the model.
- Blocking every OpenAI bot to protect content, then wondering where the referrals went. The training decision and the visibility decision are separate.
Mentioned in
Sources
Last checked against these sources on 23 September 2026.