OAI-SearchBot
The crawler behind ChatGPT's search answers. Blocking it removes you from ChatGPT citations — and it is the most damaging accidental block we see.
| Operator | OpenAI |
|---|---|
| What it does | Search & citations |
| Obeys robots.txt | Yes, per its operator's docs |
| Governs | ChatGPT Search |
| Has a crawler | Yes |
| Last verified | 13 August 2026 |
What blocking it costs you
What to know
This is the token that decides whether ChatGPT can surface and cite your site when someone asks a question you have already answered. It is separate from GPTBot, and the two are constantly confused.
The mistake has a shape: someone copies a "block AI scrapers" snippet from a blog post, the snippet lists every OpenAI token it can find, and the site quietly leaves ChatGPT's index. Nothing warns you. Blocking OAI-SearchBot while allowing GPTBot is almost always backwards from what was intended.
OpenAI's documentation notes that blocked sites may still appear as bare navigation links — a name with no content behind it, which is not the same as being recommended.
The robots.txt rules
To block OAI-SearchBot:
User-agent: OAI-SearchBot
Disallow: /To allow it explicitly — worth doing when your file also contains a blanket User-agent: * disallow, since the most specific matching group wins and a token with no group of its own falls back to the wildcard:
User-agent: OAI-SearchBot
Allow: /If neither rule is present, OAI-SearchBot is allowed. That is the default, and it is why an accidental block is nearly always something that was added rather than something that was forgotten — usually a snippet copied from a blog post that listed more tokens than the author intended to block.
Is it blocked on your site?
Your robots.txt is only the half you can read. A CDN or bot-management rule can turn OAI-SearchBot away while your file still says it is welcome, and that mismatch is invisible from the front of the site. We check both layers separately and tell you which one is doing what.
We request two things: your robots.txt and your homepage. No other pages, ever.
Official sources
- OpenAI’s crawler documentation
- Published IP ranges — the only way to verify a request claiming this user agent really came from OpenAI
Other OpenAI tokens
- GPTBot — Model training
- ChatGPT-User — Fetches when a user asks