ChatGPT-User: user agent, IP verification and robots.txt

ChatGPT-User is operated by OpenAI. It is classified here as Not a crawler (User-triggered fetch). Verification method: Published CIDR list.

Exact user-agent string

Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot

robots.txt token: ChatGPT-User. The user-agent string is self-asserted and can be forged by anyone. It is a claim, not proof.

How to verify ChatGPT-User is genuine

Match the source IP against OpenAI's published CIDR list: https://openai.com/chatgpt-user.json

322 ranges published, as of 2026-07-30. Pulled live at build time, never hardcoded.

How to block or allow ChatGPT-User

User-agent: ChatGPT-User
Disallow: /

Replace Disallow with Allow to permit it explicitly.

What blocking ChatGPT-User costs you

Blocking ChatGPT-User stops OpenAI retrieving your page when a specific user asks a question that needs it. It has no effect on training and no effect on whether you appear in ChatGPT search. You are blocking the moment of highest intent, a real person who has already been pointed at your content.

What is specific to ChatGPT-User

ChatGPT-User is not a crawler and OpenAI does not treat it as one. Because the fetch is initiated by a person, OpenAI states robots.txt rules may not apply, so a Disallow is a request rather than a control. It is also by far the largest published range set of any AI fetcher, which is what you would expect from traffic that originates wherever users are rather than from a fixed crawl fleet. That size matters operationally: an allowlist built from this file needs refreshing on a schedule, not once.

Verify a request claiming to be ChatGPT-User

Check a ChatGPT-User request against OpenAI's published ranges

Source

Primary documentation: https://platform.openai.com/docs/bots. Last verified 2026-08-02.

Other OpenAI crawlers

All crawlers and the full comparison table