OpenAILive fetcher
ChatGPT Agent
ChatGPT Agent is an OpenAI agent used when ChatGPT navigates websites for user-directed tasks.
Reads a page when someone asks ChatGPT about it.
- robots.txt token
ChatGPT Agent- Follows robots.txt
- Not documented
- How to verify it
- HTTP Message Signatures
- Seen on
- Not measured
- Last checked
- 21 July 2026
Should you block ChatGPT Agent?
Usually not. It reads a page only when a person asks ChatGPT about it, so blocking it hides your page from someone already looking for it.
| OpenAI bot | Job | If you block it |
|---|---|---|
| GPTBot | Training | Disallowing GPTBot expresses an opt-out for the potential training use covered by that crawler. ChatGPT search uses the separate OAI-SearchBot control. |
| OAI-SearchBot | Search | Disallowing OAI-SearchBot can reduce search discovery and citation eligibility. Allowing it does not guarantee indexing, ranking, or citation. |
| ChatGPT AgentThis page | User fetch | A verified request represents task automation on a user’s behalf, not background search indexing or training crawling. |
| ChatGPT-User | User fetch | A matching request is consistent with a user-triggered fetch, not broad search indexing or training. It does not reveal the conversation or prove citation. |
| OAI-AdsBot | Other | Blocking it can prevent an ad landing page from being reviewed or matched to relevant ad opportunities. It does not affect organic ChatGPT search or model training. |
From OpenAI’s documentationOpenAI says robots.txt rules may not apply to ChatGPT-User, because a person starts each request.
No rule to copy
OpenAI documents signed-request verification, but its current guidance does not state a robots.txt guarantee for ChatGPT Agent. A made-up rule could match nothing, or the wrong crawler.
ChatGPT Agent fires when a real user's question pulls in your page. To know how often ChatGPT reaches for your brand at all, track your brand mentions in ChatGPT.
Spot it in your logs.
Match the name, then check where the request came from. Anyone can copy a user-agent.
ChatGPT AgentVersion numbers change. Match the highlighted name, not the whole string.
- 1
Match the name
Search your server or CDN log for
ChatGPT Agent, case-insensitive. A match is a clue, not proof. - 2
Check the HTTP Message Signatures
OpenAI signs every outbound request. Validate the signature or use a CDN’s verified-bot detection instead of trusting the user-agent text alone. help.openai.com/en/articles/11845367-chatgpt-agent-allowlisting/
Is this really ChatGPT Agent?
Signed requestsOpenAI signs each ChatGPT Agent request with HTTP Message Signatures (RFC 9421), so its address proves nothing. A genuine request carries Signature, Signature-Input and Signature-Agent headers, and the signature checks out against OpenAI’s published keys.
Many CDNs, Cloudflare included, check these signatures for you and mark the request as a verified bot. Without a signature, it isn’t ChatGPT Agent.
help.openai.com/en/articles/11845367-chatgpt-agent-allowlistingServe the content in HTML
The cited operator material does not verify JavaScript rendering. Serve useful HTML before client-side JavaScript where possible.
OpenAI documents signed-request verification, but its current guidance does not state a robots.txt guarantee for ChatGPT Agent.
Questions, answered.
ChatGPT Agent is an OpenAI agent used when ChatGPT navigates websites for user-directed tasks.
Sources.
Officially documented. Operator facts and Trakkr’s measurements keep separate dates.
- [1]OpenAI crawler documentation
Officially documented · checked 21 July 2026
help.openai.com/en/articles/11845367-chatgpt-agent-allowlisting/
First verified 21 July 2026. Last source check 21 July 2026.
See which AI crawlers read your site.
Trakkr shows which bots, ChatGPT Agent included, visit which pages each day, and whether the answers they feed mention you.
14-day free trial. Cancel anytime.