SearchGPT
SearchGPT was the name OpenAI gave to its prototype of web search inside ChatGPT, since folded into the product as ChatGPT search. The feature answers a question in prose with links to the pages it drew on, and it is fed by a crawler of its own, OAI-SearchBot.
How a site gets into ChatGPT search
OpenAI’s crawler documentation names the agent: OAI-SearchBot surfaces websites in ChatGPT’s search features. A site that disallows it will not be shown in ChatGPT search answers, though it can still appear as a navigational link. The same page lists GPTBot, whose rule signals whether content may be used to train models, and ChatGPT-User, which fetches a page when a user asks for it. The name SearchGPT does not appear in that documentation any more; the crawler token is what a site owner needs to know.
What this site has and has not measured
A search answer is written from fetched pages, so what ChatGPT search says about a business is what the business’s own site and the pages about it serve to a crawler. This site has measured what 9,483 small-business sites serve, in what 9,483 small business websites serve to an AI crawler. It has not probed ChatGPT search itself: its one-question probe of businesses used Claude with retrieval, and a Claude number is not a ChatGPT number. There are no ChatGPT search rates on this page for that reason.
A common question
Does blocking GPTBot keep a site out of ChatGPT search?
No. OpenAI’s documentation gives GPTBot a training-data role and OAI-SearchBot the search role, with separate rules. A site can disallow GPTBot and allow OAI-SearchBot, or the reverse, and only the OAI-SearchBot rule decides whether ChatGPT search answers can show it.
Related terms
Sources
Crawler roles and the effect of disallowing them: OpenAI’s crawler documentation. The audit is this site’s own; it did not probe ChatGPT search.