GPTBot
OpenAI · reviewed 2026-11-05
About GPTBot
Collects public pages that may be used to train OpenAI models.
Operates ChatGPT and the OpenAI API; separates user-triggered fetches, search indexing, ads fetching and training crawlers into distinct user agents.
Usually means: A model provider fetched this URL for future training data. It does not mean today's answers cite the page.
Identity
User agent
Crawlers are matched by token, not the full browser string. Providers may add version or product text around it.
robots.txt token
Use this token in a User-agent line to allow or block the crawler.
Block this token
User-agent: GPTBot Disallow: /
A Disallow only stops crawlers that honor robots.txt. Classify the category before blocking; check the live file with the robots.txt checker.
Verify a hit
A matching user agent is not proof: any client can send the string. When the vendor publishes network data, compare the request against it before you trust the hit.
Match the user agent
Request UA contains GPTBot.
Check reverse DNS
The vendor publishes no reverse-DNS suffix.
Check the IP range
Fetch the endpoint below and compare the request IP.
Range endpoints
A verified hit proves the network path; it does not prove what the crawler will do with the page.
GPTBot: common questions
What is GPTBot?
GPTBot is a crawler from OpenAI, filed under Training. Collects public pages that may be used to train OpenAI models. Requests advertise it in the User-Agent header as GPTBot.
Does GPTBot train AI models?
Yes. It collects content for model training, which shapes future models rather than today's answers or citations.
How do I block GPTBot in robots.txt?
Add a group with User-agent: GPTBot and Disallow: / to robots.txt, then re-test the live file. robots.txt matches the token, not the full User-Agent string, and blocking this token does not block the vendor's other crawlers.
How do I verify a GPTBot hit?
Match the user agent, then compare the request IP with the ranges or endpoints listed on this page. A matching user agent alone is not proof — any client can send the string.
GPTBot vs ChatGPT-User: what is the difference?
ChatGPT-User is filed under AI answers, while GPTBot is filed under Training. They use separate robots.txt tokens: blocking one does not block the other, and a hit from one says nothing about the other.
Testing a block? Run the live file through the robots.txt tester or check the whole AI crawler access group.
See when GPTBot hits your site
Page scripts never see AI crawlers because they do not run JavaScript. Install the server probe and each hit lands in the Crawlers block with its path and status.