OpenAI’s documentation says it uses ChatGPT-User for certain user actions in ChatGPT and Custom GPTs, and that it is not used for crawling the web automatically. In practice, it is the visit you see when someone asks ChatGPT to look at a specific page, or when a custom GPT fetches something on a user’s behalf.
The documentation also notes that because these actions are initiated by a user, robots.txt rules may not apply. That makes ChatGPT-User different from OAI-SearchBot and GPTBot, which you control through robots.txt.
For a brand, the practical points are simple. Don’t rely on robots.txt alone to keep pages away from user-initiated visits; put anything private behind a login. And make sure your public pages show their main text in the HTML, without needing JavaScript, so a visit on a user’s behalf can actually read them.
Read can ChatGPT find your website? or test a page with the free checker.