ChatGPT-User
OpenAI · User-requested fetches
What OpenAI says it does
“certain user actions in ChatGPT and Custom GPTs”
The facts
| Operator | OpenAI |
| robots.txt token | ChatGPT-User |
| Appears in access logs | Yes |
| User-agent string | Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot |
| robots.txt | May not honor robots.txt. OpenAI states robots.txt rules may not apply to it because the fetches are initiated by a user, not a crawl schedule. |
| IP list | https://openai.com/chatgpt-user.json |
| Documentation | https://developers.openai.com/api/docs/bots, read 4 Aug 2026 |
Allow or block it in robots.txt
To block ChatGPT-User entirely:
User-agent: ChatGPT-User Disallow: /
To allow it everywhere:
User-agent: ChatGPT-User Allow: /
OpenAI's stated position: may not honor robots.txt. OpenAI states robots.txt rules may not apply to it because the fetches are initiated by a user, not a crawl schedule. Blocking it reliably means blocking at the server, not in robots.txt.
Verify a request is really ChatGPT-User
Match the requesting IP against the published list.
A hit from ChatGPT-User means a person asked ChatGPT something and it opened this page to answer. That is exactly the event AI visibility tools try to measure.
Observed on this property
Not observed here yet. This site logs every request from 15 named AI agents; logging began 4 Aug 2026. When ChatGPT-User first fetches a page, this section fills in from the log.
Also run by OpenAI
- GPTBot | Trains AI models
- OAI-SearchBot | AI search index