# ChatGPT Agent

> ChatGPT's agent mode driving a real browser: it navigates and interacts with sites to finish a multi-step task a user gave it. OpenAI governs it with the ChatGPT-User token and the ChatGPT-User prefix list rather than a token of its own, so the robots rule and the address check are the same ones.

| field | value |
|---|---|
| operator | OpenAI |
| category | User-triggered fetchers |
| robots.txt token | `ChatGPT-User` |
| user-agent contains | `ChatGPT Agent` |
| robots.txt | obeys robots.txt (documented) |
| verify by | published IP ranges |
| published IP ranges | https://openai.com/chatgpt-user.json |
| prefixes mirrored | 204 IPv4 / 0 IPv6 |
| operator docs | https://platform.openai.com/docs/bots |
| last reviewed | 2026-09-01 |

## What blocking it costs you

Agentic tasks a user asked for — booking, comparing, filling a form on your site — fail. This is the fetch that ends in a transaction, so it is the most expensive user-triggered block on this list.

## Full user-agent string

```
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot
```

## Block it

```
User-agent: ChatGPT-User
Disallow: /
```

## Allow it

```
User-agent: ChatGPT-User
Allow: /
```

JSON: https://www.pathwren.workers.dev/crawler/chatgpt-agent.json · index: https://www.pathwren.workers.dev/llms.txt