Training / data-use control 4 min read  ·  Updated October 2026

GPTBot: purpose, robots.txt rules and access checks

GPTBot collects pages that may be used to train OpenAI models. Blocking GPTBot expresses a training opt-out. ChatGPT search has its own OAI-SearchBot control.

maintains Let The Bots In, a technical AI-readiness scanner and reference site.

Technical reference

OperatorOpenAI
ClassTraining / data-use control
HTTP User-AgentMozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.4; +https://openai.com/gptbot
Citation impactBlocking GPTBot expresses a training opt-out. ChatGPT search has its own OAI-SearchBot control.
IP verificationOfficial IP ranges or verification reference

Header versions can change. Use the product token for robots.txt matching and consult the OpenAI crawler documentation before changing enforcement. Reference checked on 18 October 2026.

How do I allow GPTBot?

Merge this group into the robots.txt served from the page's origin. Keep any path exclusions that still apply. A named group may stop inheriting the wildcard group's exclusions.

User-agent: GPTBot
Allow: /

Allow retrieval and block training

This example separates the operator's documented identities where possible. It changes crawl or data-use preferences; it does not grant server access. User-triggered agents may follow a different robots.txt policy.

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: GPTBot
Disallow: /

Use the interactive robots.txt builder for a per-bot policy. Test the final file against the exact page path before deploying it.

What if the CDN returns HTTP 403?

A robots.txt allow rule cannot override an HTTP 403, a login gate or a JavaScript challenge. Compare the returned body with the public page and locate the request in CDN security events. Grant an exception only after verifying the sender, scoped to the affected public routes.

The scanner uses its own network. A successful header probe does not verify OpenAI access, and a refused probe does not establish a vendor-wide block. See the WAF troubleshooting guide.

Check your site's rules for GPTBot

Frequently asked questions

What does GPTBot do?

OpenAI operates GPTBot. GPTBot collects pages that may be used to train OpenAI models.

Does blocking GPTBot affect citations?

Blocking GPTBot expresses a training opt-out. ChatGPT search has its own OAI-SearchBot control.

How do I verify a GPTBot request?

Use the operator's published verification reference linked on this page. Match the source IP, not just the User-Agent header. A scanner probe comes from its own network.

Check your site's technical readiness

Scan a public URL for crawler rules, delivered HTML and page metadata. The report distinguishes measured issues from checks it could not complete.

Check my site →