# GPTBot

> GPTBot is OpenAI's crawler for collecting pages that may be used to train AI models. Here's how to recognize it and check a request really is it.

## GPTBot facts

- **Operator:** OpenAI
- **Type:** Crawler
- **Why it fetches:** Collects training data
- **User agent contains:** `GPTBot`
- **Verified by:** published IP ranges
- **Operator's documentation:** [platform.openai.com/docs/bots](https://platform.openai.com/docs/bots)

## How to tell a real GPTBot from an impostor

Anyone can put "GPTBot" in a user agent. What can't be faked is below.

- **Check the address.** OpenAI publishes the IP ranges GPTBot uses: [openai.com/gptbot.json](https://openai.com/gptbot.json). A request whose user agent says GPTBot from an address outside them is an impostor.

## How Punchline counts GPTBot

Punchline labels a request that names GPTBot with how sure it is: **verified** or **declared** when the evidence checks out, a lower tier when it can't be checked, and **impostor** when the evidence contradicts it. Each visit is counted under OpenAI, with its purpose, so you can see whether GPTBot is reading your site and which pages it reads. [How it works](https://punchlineanalytics.com/how-it-works) · [Known agents](https://punchlineanalytics.com/docs/agents)

[See which agents read your site](https://punchlineanalytics.com/reveal)
