GPTBot Analytics
OpenAI's crawler reads your site more often than you think — and no analytics tool shows it, because GPTBot never runs JavaScript.
This page is the reference: what GPTBot is, how it differs from OpenAI's other two bots, how to track it live, and how to block it without vanishing from ChatGPT's answers.
OpenAI runs three bots. They mean different things.
Most robots.txt advice treats them as one. They're not — and the difference decides whether "blocking OpenAI" costs you customers.
GPTBot
TrainingCollects public web content to train OpenAI's models. Blocking it costs you no visibility in ChatGPT's answers — training and answering are separate pipelines.
OAI-SearchBot
Search indexBuilds the index ChatGPT Search draws on. This is the one you almost always want to ALLOW — it's how you appear in AI answers.
ChatGPT-User
Live user requestFetches your page because a ChatGPT user is asking about it right now. A hit from this bot means a real person, mid-conversation, about you.
robots.txt — block training, keep the answers
User-agent: GPTBot
Disallow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /Questions
- What is GPTBot's user-agent string?
- Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.2; +https://openai.com/gptbot — the stable identifier is the "GPTBot" token. OpenAI also publishes its IP ranges for verification.
- Why can't I see GPTBot in Google Analytics?
- Google Analytics measures visitors by running JavaScript in their browser. GPTBot — like every AI crawler — fetches your HTML and never executes the tracking script, so it's invisible there. Seeing it requires something server-side: a middleware, a Worker, or log analysis.
- How do I block GPTBot?
- Two lines in robots.txt: "User-agent: GPTBot" then "Disallow: /". OpenAI honors it. Important: block GPTBot alone if your concern is training — leave OAI-SearchBot and ChatGPT-User allowed, or you'll also disappear from ChatGPT's answers.
- Should I block GPTBot?
- It's a genuine trade-off. Blocking stops your content training future models; allowing it may make those models understand your product better. What's not a trade-off: the decision should be explicit. Most sites have simply never chosen — our scanner flags that.
- How often does GPTBot crawl?
- It's bursty: days of silence, then a sweep of many pages. Frequency correlates loosely with how often your content changes and how linked it is. The only way to know your actual pattern is to measure it — which is what the live tracker below does.
Stop guessing what GPTBot reads. Watch it.
2-minute install · humans never tracked · free while in beta
Set up GPTBot tracking →