GPTBot: OpenAI's training crawler
GPTBot is OpenAI's web crawler tasked with collecting content to train its foundation models (the GPT family). It does not answer ChatGPT user queries: that role belongs to OAI-SearchBot (search index) and ChatGPT-User (live browsing). Blocking GPTBot removes your content from future training corpora, with no documented effect on your visibility in ChatGPT search.
Technical profile
- Operator
- OpenAI
- Role
- Training crawler
- robots.txt token
GPTBot- User-agent (example)
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.1; +https://openai.com/gptbot- Respects robots.txt
- Yes (documented by OpenAI)
- Official IP ranges
- 336 ranges published by the operator across all its bots (snapshot of 2026-07-23, refreshed on every deployment) — official file
- Official documentation
- OpenAI bots documentation
What GPTBot actually does
Every page GPTBot downloads may feed the datasets used to train OpenAI's next models. It is a periodic, large-scale crawl, mechanically similar to a search engine's — but its purpose differs: it doesn't build an index that sends you traffic, it builds a corpus.
That's the key distinction for your GEO decisions: allowing GPTBot means accepting that your content contributes to the “internal” knowledge of future models (which may restate its substance without citing you); blocking it removes you neither from ChatGPT search nor from cited answers, which rely on other agents.
GPTBot in your server logs
In an access.log file, GPTBot is identified by the “GPTBot” substring in the user-agent. Its crawl is usually moderate in volume and focused on content pages (articles, guides) rather than technical resources. Regular GPTBot visits indicate your site is considered a useful content source for training.
Verifying it's really GPTBot
A user-agent alone is trivially forged: any scraper can claim to be “GPTBot”. OpenAI publishes the official list of its IP ranges in a dedicated JSON file — the only reliable method is checking that each hit's IP belongs to those ranges. That's exactly what our log analyzer does, automatically, without your files ever leaving your browser.
View GPTBot's official IP ranges (JSON)Blocking or allowing GPTBot (robots.txt)
GPTBot honors robots.txt directives according to OpenAI's documentation. Blocking it is a legitimate editorial choice (a training opt-out) that affects neither Googlebot nor ChatGPT search. Don't conflate agents: blocking GPTBot blocks neither OAI-SearchBot nor ChatGPT-User, which have their own tokens.
To block it:
User-agent: GPTBot Disallow: /
To explicitly allow it:
User-agent: GPTBot Allow: /
Frequently asked questions
Should I block GPTBot?
It's an editorial trade-off, not a technical one. Blocking GPTBot removes your content from OpenAI's future training corpora — useful if you refuse that reuse. It does not remove your visibility in ChatGPT search (OAI-SearchBot) or live page browsing (ChatGPT-User). Many visibility-focused sites choose to allow everything.
Does GPTBot respect robots.txt?
Yes. OpenAI publicly documents that GPTBot reads and honors robots.txt, and publishes its IP ranges to allow verification. Third-party tests confirm this behavior. A “GPTBot” ignoring your directives is most likely an impostor — verify it by IP.
How do I know if GPTBot visits my site?
Only your server logs show it: GPTBot doesn't execute JavaScript, so it's invisible to Google Analytics and tag-based tools. Drop your access.log files into our analyzer: it isolates GPTBot, verifies its authenticity by IP and lists the pages it requests.
Does blocking GPTBot remove my site from ChatGPT?
No. ChatGPT answers that cite websites rely on OAI-SearchBot's index and on ChatGPT-User's live fetches. GPTBot only serves model training. They are three distinct agents with three distinct robots.txt tokens.
Sources
Every fact on this page is backed by a primary source — the operator's official documentation and, where it exists, its published IP list.
See what this bot does on your site
Drop your access.log files into the Crawl Lab log analyzer: it isolates GPTBot, verifies its authenticity and lists the pages it requests — entirely in your browser, no upload.
Analyze my logs