OpenAI: robots.txt may not apply to ChatGPT-User
If you blocked ChatGPT-User in robots.txt to keep ChatGPT away, that may no longer work. Here is how OpenAI’s three bots are documented now.
OpenAI’s crawler documentation now says that robots.txt rules may not apply to ChatGPT-User, the agent that fetches a page when a ChatGPT user asks about it. Earlier versions described it as controllable through robots.txt.
What happened
OpenAI documents three agents:
- GPTBot collects content that may be used to train OpenAI’s models. robots.txt applies.
- OAI-SearchBot crawls for ChatGPT search. OpenAI points site owners to this bot to manage whether pages appear in search answers.
- ChatGPT-User visits pages on behalf of a user, for example when someone pastes a link or ChatGPT browses during an answer. These fetches are treated like a user’s request.
This is consistent with how Google treats user-triggered fetchers. One source also reports newer version numbers in GPTBot and OAI-SearchBot user agents. That is a single report.
What it means for you
- To stay out of training, disallow
GPTBot. - To stay in ChatGPT search, allow
OAI-SearchBot. - To stop ChatGPT-User, robots.txt may not be enough. A firewall rule is the only hard block, and it also stops users who share your links in ChatGPT.
Check how often each OpenAI bot visits in your logs with the AI Bot Log Analyzer.