ChatGPT-User user agent, IP ranges and robots.txt

OpenAI operates ChatGPT-User, a fetcher that may visit a page when a ChatGPT or Custom GPT user asks a question, and notes ChatGPT users may reach external applications via GPT Actions. OpenAI states it does no automatic crawling. Because a user initiates each request, robots.txt rules may not apply; no robots.txt token is named. Verification uses OpenAI's IP range file.

Logs show a user-agent string that OpenAI labels as the full string: Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot. Each request corresponds to a person asking ChatGPT or a Custom GPT something that leads the assistant to open a page. OpenAI also mentions, under the same agent, that ChatGPT users may interact with external applications via GPT Actions. OpenAI is explicit that ChatGPT-User is not used for crawling the web in an automatic fashion, so a site owner should expect fetches tied to user prompts rather than systematic traversal of a site.

The category is user-triggered fetch, and OpenAI's documentation states that because these actions are initiated by a user, robots.txt rules may not apply. The docs also say ChatGPT-User is not used to decide whether content may appear in Search; OAI-SearchBot is the robots.txt token for managing Search opt-outs and automatic crawl, and GPTBot is the other token the page names. OpenAI does not present ChatGPT-User itself as a robots.txt control token. Blocking it at the server level stops ChatGPT from reading pages on request, which affects on-demand answers rather than search inclusion.

For verification OpenAI publishes chatgpt-user.json at openai.com. The file holds a creationTime and a prefixes array of ipv4Prefix entries. OpenAI documents no reverse-DNS method for this agent and says nothing about whether it executes JavaScript, so both remain undocumented. Refresh the file on a schedule; the creationTime field records when OpenAI generated the copy you hold, and an address outside the current list is not the documented agent.

Operator note. The documentation states: "Because these actions are initiated by a user, robots.txt rules may not apply." It also states: "ChatGPT-User is not used to determine whether content may appear in Search. Please use OAI-SearchBot in robots.txt for managing Search opt outs and automatic crawl." The documentation does not note that the version number may change, and it does not present ChatGPT-User as a robots.txt control token; the only robots.txt tags it names are OAI-SearchBot and GPTBot. The documentation does not say whether ChatGPT-User executes JavaScript.