ChatGPT-User robots.txt group (may not apply)
What it checks
Whether robots.txt has a group that names ChatGPT-User. A catch-all group that happens to cover the token is not reported: this audit speaks only about a group the owner wrote for ChatGPT-User.
| State | Result |
|---|---|
a User-agent: ChatGPT-User group, allowing or blocking |
pass |
| no robots.txt, a soft 404, or no group naming ChatGPT-User | notApplicable |
The pass message quotes OpenAI: the group may not be honoured, and OAI-SearchBot is the token for search opt-outs. Matching is case-insensitive and ignores a version suffix such as ChatGPT-User/1.0.
Why it matters
Mechanism: a named robots.txt group controls the crawler that sends that product token. The claim this audit tests is that OpenAI does not commit to obeying the group for ChatGPT-User, so it is not a reliable control.
Evidence
- OpenAI: “ChatGPT-User is not used for crawling the web in an automatic fashion. Because these actions are initiated by a user, robots.txt rules may not apply. … Please use OAI-SearchBot in robots.txt for managing Search opt outs and automatic crawl.” — Overview of OpenAI Crawlers (verified 2026-10-18)
Limits
“May not apply” is not “never applies”. The audit says the group may not be honoured, not that it is ignored.
How it scores
Grade: C — OpenAI says robots.txt rules “may not apply” to ChatGPT-User. That is a hedge, not a documented refusal, so the group is informative with an unconsumed notice rather than obsolete (docs/evidence/policy.md, “Informative or obsolete”). The audit carries weight 0, never warns or fails, and does not move any score.
Example finding
robots.txt has User-agent: ChatGPT-User / Disallow: /. The report says the rule may not be honoured and names OAI-SearchBot for search opt-outs.
Sources
- OpenAI Bots / Crawler documentation — OpenAI, vendor-doc (verified 2026-10-19)
- RFC 9309: Robots Exclusion Protocol — IETF, spec (verified 2026-08-21)
- OpenAI crawlers and user agents — OpenAI, vendor-doc (verified 2026-10-19)
- A deeper look at AI crawlers: breaking down traffic by purpose and industry — Cloudflare, article (verified 2026-08-20)
- 15% of AI page fetchers in Europe reached disallowed URLs, TollBit finds — PPC Land, reporting TollBit ‘State of the Bots’ H1 2026 (‘The Pipes are Leaky’), study (verified 2026-08-20)
- The Paradox of Blocking AI Crawlers: You Lose Visitors, Not Citations — Pebblous, synthesizing BuzzStream/XOFU and Zhao (Rutgers) & Berman (Wharton), study (verified 2026-08-20)