ChatGPT-User
User-Agent token: ChatGPT-User · Operator: OpenAI · Category: AI Assistant · User-Triggered Fetcher
FraudMeNot Verdict
Verified AI Agent (User-Triggered)
Published by OpenAI. Visits a page only when a ChatGPT user asks it to; not used for model training per OpenAI.
Can Be Spoofed?
Yes — User-Agent Header Spoofable
Anyone can copy the header. Verify the source IP against OpenAI's 235 published CIDR blocks in chatgpt-user.json.
Abuse & Unwanted Activity Reports
No Public Abuse Flags
User-triggered fetches may ignore robots.txt, per OpenAI. No CVE names this fetcher itself.
Judgment Call · What You Should Do
Generally Legitimate — But Verify the IP, and Never Give Any AI Agent Your Bank Password
For everyday users: If you use ChatGPT to read a web page for you, this is the bot that fetches it. That is normal. But do not paste your bank, brokerage, or email username, password, or one-time codes into any AI assistant. In October 2026 Axios reported that major U.S. banks are detecting and blocking third-party AI agents that log into customer accounts, and that letting an agent use your login goes against most banks' rules.
For website owners & security teams: Allow or rate-limit it on public content pages if you want to appear in ChatGPT answers. Treat any request that claims to be ChatGPT-User from an IP outside OpenAI's published list as spoofed. Keep login, checkout, and money-movement endpoints behind human verification regardless of User-Agent.
Report & Activity Summary
235
Published IP Ranges (CIDRs)
69,280 IPv4 addresses · operator JSON feed
1
Known User-Agent Signatures
crawler-user-agents (MIT) dataset
Partially
robots.txt Compliance
ai.robots.txt registry (MIT)
0
FraudMeNot Tallies
Humans: 0 · AI Agents: 0
For AI Agents & Automated Gatekeepers
- Agent_Identifier
ChatGPT-User- FraudMeNot_Verdict
- Verified AI Agent (User-Triggered)
- Operator_Entity
- OpenAI
- Agent_Category
- AI Assistant · User-Triggered Fetcher
- User_Agent_Spoofable
true- Verification_Method
- IP CIDR match against
https://openai.com/chatgpt-user.json· Web Bot Auth: Not documented for this fetcher. OpenAI says a separate agent, ChatGPT Work's Cloud browser, signs its requests with HTTP Message Signatures (RFC 9421). Its requests carry a Signature-Agent header set to "https://chatgpt.com", and Cloudflare verifies these signatures automatically. - Respects_Robots_Txt
- Partially (user-triggered fetches may ignore robots.txt, per OpenAI)
- Auth_And_Banking_Policy
- Block credentialed logins and money movement from any AI agent; allow read-only via authorized open-banking APIs.
- Recommended_Gatekeeper_Action
- ALLOW public reads from verified OpenAI IPs; CHALLENGE or BLOCK when IP does not match chatgpt-user.json; always require human step-up on /login and payment routes.
- Canonical_Citation_URL
- https://fraudmenot.com/bot/chatgpt-user/
Technical Identity, IP Verification & Spoofing Check
| Full User-Agent string(s) |
|
|---|---|
| robots.txt token | ChatGPT-User |
| Operator | OpenAI · Official bot documentation |
| Official IP verification feed | https://openai.com/chatgpt-user.json — 235 CIDR blocks (69,280 IPv4 addresses), file dated 2026-10-27 |
| Sample published CIDR blocks | 104.208.184.192/28, 104.210.139.192/28, 104.210.139.224/28, 128.85.198.32/28, 13.65.138.112/28, 13.67.72.16/28 |
| Reverse DNS (rDNS) | No rDNS hostname is documented; verify by IP list (CIDR) match. |
| Web Bot Auth (Signature-Agent) | Not documented for this fetcher. OpenAI says a separate agent, ChatGPT Work's Cloud browser, signs its requests with HTTP Message Signatures (RFC 9421). Its requests carry a Signature-Agent header set to "https://chatgpt.com", and Cloudflare verifies these signatures automatically. |
| robots.txt behavior | OpenAI says robots.txt rules for ChatGPT-User may not apply in the same way because each fetch is initiated by a user; block it at the WAF if you need a hard stop. |
Copy-Paste robots.txt & Cloudflare WAF Rules for ChatGPT-User
robots.txt — allow public pages, keep private routes off-limits
User-agent: ChatGPT-User
Allow: /
Disallow: /login/
Disallow: /account/
Disallow: /checkout/robots.txt — block completely
User-agent: ChatGPT-User
Disallow: /Cloudflare WAF custom rule expression
(http.user_agent contains "ChatGPT-User" and not cf.client.bot)Blocks requests that claim to be ChatGPT-User but are NOT recognized by Cloudflare as a verified bot (i.e., likely spoofed). Action: Block or Managed Challenge.
Capabilities, Data Usage & Autonomous Action Risk
| Trains AI on page content? | No (OpenAI says ChatGPT-User content is not used to train models) |
|---|---|
| Executes JavaScript? | Fetches pages on demand; agent-mode browsing can render JavaScript |
| Can submit forms or attempt logins? | Standard fetches: No. ChatGPT agent mode (a separate signed browser): can click, fill forms, and attempt logins when the user asks. |
| Banks & login pages should | Block credentialed logins and money movement from any AI agent; allow read-only via authorized open-banking APIs. |
| E-commerce checkouts should | Allow product page reads; require human confirmation (e.g., step-up or CAPTCHA) for checkout. |
Context: In October 2026, Axios reported that major U.S. banks were detecting unvetted third-party AI agents inside customer accounts, and CNBC reported that JPMorgan blocks outside AI agents from logging in with customers' credentials while allowing read-only data sharing through open banking providers.
Seen ChatGPT-User? Quick Closed-Choice Tally
Choose from preset options only. No free-text comments, so nobody can post names, profanity, or unverified accusations.
Common Questions About ChatGPT-User
What is ChatGPT-User?
ChatGPT-User is the User-Agent OpenAI uses when a ChatGPT user asks ChatGPT to open a specific web page or when a Custom GPT calls a web action. It is user-triggered, not an automatic crawler.
Is ChatGPT-User safe to allow on my website?
For public pages, it is generally a legitimate visitor operated by OpenAI. Because the User-Agent header can be forged, compare the source IP against OpenAI's 235 published CIDR blocks at openai.com/chatgpt-user.json before trusting it.
Should I let an AI agent log into my bank account?
FraudMeNot recommends against giving any AI agent your bank username, password, or one-time codes. Major U.S. banks reported in October 2026 that they are blocking third-party AI agents from logging into customer accounts; most support read-only data sharing through authorized providers instead.
Could This Be a Scam? What to Do
- Anyone can copy the ChatGPT-User User-Agent. OpenAI publishes an IP list for it, so compare the visiting address against that list before you trust the visit.
- Never give an AI agent your logins, bank access, or one-time codes unless you set it up yourself.
- If a call, text or email asking for money or access uses the ChatGPT-User or OpenAI name, stop and report it.
Guide: how to check whether a crawler visit is really who it claims to be.
Compare OpenAI's Crawlers
GPTBot vs OAI-SearchBot vs ChatGPT-User: what each does, a side-by-side table with robots.txt examples.
More from OpenAI
- ChatGPT Agent — AI Agents
- GPTBot — AI Training Crawler
- OAI-AdsBot — Validates and targets ads on ChatGPT
- OAI-SearchBot — AI Search Crawler (ChatGPT search)
- AddSearchBot — AI Search Crawlers
- adidxbot — Advertising quality crawler
Related on FraudMeNot
- Verified security vulnerabilities, explained
- Report a scam, or find where to report it in your state
- Browse the AI Agent & Bot Directory
Sources (checked Nov. 16, 2026): ai.robots.txt (MIT), crawler-user-agents (MIT), OpenAI IP feed, Axios (Oct. 22, 2026), CNBC (Oct. 13, 2026), OpenAI Help Center: Web Bot Auth. FraudMeNot is not affiliated with OpenAI. Product and company names are trademarks of their owners and are used here only to identify the bots. General information, not security advice.