GPTBot: OpenAI's Training Crawler
What is GPTBot?
GPTBot crawls content that may be used to train OpenAI foundation models. Blocking it signals that your content should not be used in training. GPTBot is separate from OAI-SearchBot, so blocking GPTBot keeps you out of training runs while leaving ChatGPT search answers untouched.
Key facts
- GPTBot is operated by OpenAI, and its purpose is training: collects content that may be used to train future models.
- The user agent for GPTBot is documented by OpenAI and contains the token GPTBot.
- The robots.txt token for GPTBot is GPTBot, and OpenAI documents that GPTBot respects that rule.
- GPTBot never sends a visitor to your site, and GPTBot attaches no citation when it uses your content.
Specification
| Operator | OpenAI |
|---|---|
| Purpose | TrainingCollects content that may be used to train future models. Blocking it removes your content from future training runs. It does not affect whether a model can find and cite you today. |
| Agent kind | CrawlerAutonomous crawler. Appears in your logs under its own user-agent. |
| User agent | Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.4; +https://openai.com/gptbot |
| robots.txt token | GPTBot |
| robots.txt compliance | Documented: respects robots.txt |
| Verification | Published IP range filehttps://openai.com/gptbot.json |
| Sends referral traffic | No |
| Attaches a citation | No |
| Status | Active |
| Legacy Kymo category | training |
| Source | Operator documentationhttps://developers.openai.com/api/docs/bots |
| Verified on |
What this looks like in your logs
203.0.113.42 - - [04/Aug/2026:09:14:07 +0000] "GET /pricing HTTP/1.1" 200 18432 "-" "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.4; +https://openai.com/gptbot"
The part that identifies the request is the token GPTBot, inside the quoted user-agent field at the end of the line. Everything before it is the ordinary shape of a Common Log Format entry, and the address shown is a documentation address rather than one OpenAI uses.
What GPTBot does not do
GPTBot does not send you a visitor, and it attaches no citation, so a read by it can never turn into a click.
GPTBot does not decide whether OpenAI can find you today, because a training run happens long before any question is asked.
OpenAI publishes no crawl frequency for GPTBot, so any schedule you read elsewhere is somebody's estimate rather than a documented figure.
How Kymo classifies it
Kymo files GPTBot under OpenAI as training, classified from the HTTP request on the server rather than from anything running in a browser.
A hit from GPTBot appears on the Crawlers dashboard and is deliberately left out of the AI Visibility numbers. A training run sends nobody, so a read by it has no click to compare against.
Seen on our own sites
| Last seen | 14 Sep 2026 |
|---|---|
| Days active, last 30 days | 13 |
| Requests, last 30 days | 26,307 |
Counted by Kymo on kymo.in and remotestack.in, the two sites we run, up to 15 Sep 2026. Two sites is a small sample. Treat these numbers as proof the bot is active now.
Allow it or block it
Blocking GPTBot keeps your content out of future OpenAI training runs. It does not remove you from ChatGPT search answers, which are governed by OAI-SearchBot. OpenAI notes that if both are allowed, one crawl may serve both purposes.
| Allow it if | Block it if |
|---|---|
| You want your content represented in future OpenAI models, and you accept that a training run sends no visitor and no link back. | You sell access to your content, or you object to it being used as training data without payment or credit. |
robots.txt directives
Both blocks below address GPTBot only. Rules for one token never apply to another, even from the same operator.
Allow GPTBot
User-agent: GPTBot Allow: /
Block GPTBot
User-agent: GPTBot Disallow: /
Questions
How do I know a request claiming to be GPTBot is genuine?
OpenAI publishes the IP ranges GPTBot uses at https://openai.com/gptbot.json. Check the requesting address against that file. A user-agent string alone proves nothing, because anyone can send one.
Does blocking GPTBot also block OAI-SearchBot?
No. GPTBot and OAI-SearchBot are separate tokens, and OpenAI treats each one on its own. A rule naming GPTBot leaves OAI-SearchBot free to carry on indexing your pages for OpenAI answers. To stop that as well, name OAI-SearchBot in a rule of its own.
Will I see GPTBot in Google Analytics?
No. Google Analytics runs a JavaScript tag in a visitor's browser, and GPTBot reads your HTML and leaves without running any script. Server-side logging is the only way to record it.
Does GPTBot send traffic back to my site?
No. GPTBot takes content and sends nothing back. A hit appears in your logs as a request with no visitor behind it.
Related crawlers
All AI crawlers and control tokens
See this bot in your own logs
GPTBot takes your content and sends nothing back, so it leaves no trace at all in a browser-based analytics tool.
Kymo reads the HTTP request on your server, so a crawler that never runs JavaScript is still recorded. You see which bots reached the site, which pages they took and how often, next to your human traffic.
Start free → 14-day free trial. No card required.
No account yet? Run a free AI visibility check on your own site.