GPTBot
OpenAI's crawler that collects web content potentially used to train the foundation models (ChatGPT). It does not answer a user in real time.
- Operator
- OpenAI
- User agent
-
GPTBotGPTBot/1.4 - robots.txt token
GPTBot- Respects robots.txt
- Yes
robots.txt
Respects robots.txt. A Disallow: GPTBot signals that content must not be used for training. Distinct from OAI-SearchBot (indexing for ChatGPT search) and ChatGPT-User (user-triggered fetch).
Verify authenticity
Published IP ranges
OpenAI publishes GPTBot's IP ranges as JSON (openai.com/gptbot.json). Verification means matching the source IP against that list; the user-agent alone proves nothing.
SEO recommendation
Block by default
Block by default: GPTBot feeds model training, with no citation or referral traffic. Allowing it only makes sense as a deliberate LLM-contribution strategy. Blocking it affects neither Search indexing nor visibility in ChatGPT search (handled by OAI-SearchBot).
Good to know
Typical string: Mozilla/5.0 ... compatible; GPTBot/1.4; +https://openai.com/gptbot. OpenAI's bot documentation has moved to developers.openai.com.
Official sources
Fact sheet updated on September 30, 2026