Checks / AI crawler access
robots.txt keeps GPTBot out of your home page and most of the site
GPTBot collects training data and may read only the paths robots.txt allows for it. This is a choice: most of your pages stay out of future model training, not out of AI search.
robots_partial_gptbot · AI crawler access · Notice
What triggers it
robots.txt does not let GPTBot fetch your home page, but leaves some other paths open for it; the finding names the first three.
What to change
Remove the Disallow: / rule for GPTBot in robots.txt, or add Allow rules for the other paths you want read.
Example
Lines that allow GPTBot in robots.txt:
User-agent: GPTBot
Allow: /How to confirm the fix
- Open https://example.com/robots.txt: the group that names GPTBot (or the * group, if none does) leaves most of the site open, the home page included
- Run a new crawl: the AI crawler table no longer reports GPTBot as mostly blocked
About GPTBot
GPTBot is OpenAI's training crawler. Disallowing it in robots.txt tells OpenAI the site's content should not be used to train its generative AI foundation models. It does not control ChatGPT search, which uses OAI-SearchBot. OpenAI publishes 18 IP ranges for GPTBot.
Allow it and disallow GPTBot to stay in ChatGPT search while opting out of training.
The GPTBot list at openai.com/gptbot.json is the shortest of the three OpenAI files we counted: 18 prefixes on 21 October 2026, creation time 2026-09-22. When a request names GPTBot but comes from an address outside that list, the name alone proves nothing.
Job: Training. Follows robots.txt: Yes, it follows robots.txt.
GPTBot: operator, user-agent string and address list
How common it is
We have not measured how common this is on a sample of sites. Counts appear here when a study covers it.
About this check
| Check ID | robots_partial_gptbot |
|---|---|
| Report section | AI crawler access |
| Severity | Notice |
| How it is decided | Measured: read directly from the page or the server's response. |
| Added | Before 19 Oct 2026 (versions 1 and 2), or a site-wide check with no version date yet |
| Status | Measured in every report |
| Report gives the code | Yes, one block for the site |
Settings by platform
For robots.txt, where the setting lives differs by platform. Our guides cover 13 of them: AI crawler settings by platform.
Related checks
- robots.txt keeps ClaudeBot out of your home page and most of the site
- robots.txt keeps Google-Extended out of your home page and most of the site
- Your firewall or CDN refuses AI training crawlers
- Site is not served over HTTPS
See whether it applies to your site
Paste your address and we crawl up to 10 pages and show every issue we find, with its fix. No account, no card.
Every plan crawls up to its page limit and lists these checks with the fix: plans.
How an audit works: method