OAI-SearchBot: OpenAI's other crawler, explained
OpenAI runs more than one crawler, and conflating them leads to the wrong robots.txt decision. OAI-SearchBot is the one behind ChatGPT's live web answers - not the one feeding model training.
In short
OAI-SearchBot is OpenAI's retrieval crawler: it fetches pages to support ChatGPT's search/browsing features and the live citations that go with them. It is operated separately from GPTBot, which crawls for model-training corpora. Blocking OAI-SearchBot removes you from the pool of pages ChatGPT can cite when it browses; blocking GPTBot is a narrower, training-data-only decision.
What it is and who runs it
OAI-SearchBot is OpenAI's crawler for search and browsing retrieval - the fetches behind the citations ChatGPT shows when it looks something up live, as distinct from the separate ChatGPT-User agent that fires from a specific user action inside a chat.
OAI-SearchBot vs GPTBot - the distinction that matters
| Bot | Operator | Feeds |
|---|
| GPTBot | OpenAI | Model training corpus |
| OAI-SearchBot | OpenAI | Search/browsing retrieval - live, citable answers |
Same operator, different jobs. A robots.txt rule that treats them identically is making both decisions at once, whether that was the intent or not.
Should you block it
If being citable in ChatGPT's live browsing answers is worth anything to your business, allow it. Blocking it is a narrower, defensible choice mainly for content businesses actively restricting retrieval use - most service businesses have little to gain and real reach to lose by blocking it.
How to verify it in your logs
Match the user-agent string, then cross-check the requesting IP against OpenAI's currently published ranges before trusting the label - user-agent alone can be spoofed by anyone. Verify against the operator's current documentation rather than any fixed list, including this one, since bot behavior and publishing details change.
Common questions
Is OAI-SearchBot the same as GPTBot?
No. GPTBot crawls for training data; OAI-SearchBot crawls for ChatGPT's live search and browsing retrieval. Both are operated by OpenAI but serve different jobs and should be governed by separate robots.txt rules.
Will blocking OAI-SearchBot get me removed from ChatGPT answers?
It removes your pages from that specific retrieval pathway. If OAI-SearchBot cannot fetch you, it cannot cite you when ChatGPT browses for a live answer.
How do I verify a request is really from OAI-SearchBot?
Match the user-agent string and cross-check against OpenAI's currently published IP ranges - user-agent alone can be spoofed.
Where does this fit in a broader crawler policy?
As one line in a deliberate, per-bot robots.txt policy, not a standalone decision - see the crawler policy guide for the full framework.