ChatGPT-User: when a human reads your page through ChatGPT
ChatGPT-User is not a classic crawler: it doesn't methodically walk the web. It fetches a specific page the moment a ChatGPT user needs it — to summarize an article, check a fact or cite a source in the conversation. Every ChatGPT-User hit in your logs therefore corresponds to a human reading your content live, through ChatGPT.
Technical profile
- Operator
- OpenAI
- Role
- Live browsing (user-triggered)
- robots.txt token
ChatGPT-User- User-agent (example)
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot- Respects robots.txt
- Yes (documented by OpenAI)
- Official IP ranges
- 336 ranges published by the operator across all its bots (snapshot of 2026-07-29, refreshed on every deployment) — official file
- Official documentation
- OpenAI bots documentation
Why these hits are gold
An index crawler visits thousands of pages “just in case”; ChatGPT-User only visits pages someone needs right now. It's the GEO equivalent of an organic click: real intent, a real read, often followed by a linked citation in the answer. Tracking these hits page by page shows you which content ChatGPT users actually consume.
This traffic is invisible to Google Analytics: the agent doesn't execute JavaScript tags. Your server logs are the only place to measure it — one of the most common blind spots of audience measurement in the AI assistant era.
What volume to expect
Unlike GPTBot, ChatGPT-User volume is directly proportional to user interest in your content. A few hits per week on a young site is already a positive signal: your content is surfacing in conversations. Steady growth in these hits is one of the best leading indicators of a GEO strategy that works.
What our logs show: 52 requests from 50 different IP addresses
ChatGPT-User issued 52 requests over our period, but spread across only three days and, above all, from fifty distinct IP addresses. That ratio of nearly one address per request is the signature of a user-triggered fetch: each request starts from a different context. A real crawler does the opposite — GoogleOther produced 835 requests from just ten addresses.
The browsing profile confirms it: only twelve distinct URLs, tightly concentrated, including seventeen visits to the homepage and a run of visits to pages in this section. No request to robots.txt, none to the sitemaps: this agent is not building an index, it is fetching a specific page to answer a question.
That is why you should never add ChatGPT-User to GPTBot in a crawl dashboard. The first measures user interest at a point in time, the second measures a scheduled collection. Merging them confuses a visit with a subscription.
Measured on our own server logs, 20–27 July 2026 (1,722 crawler requests kept after verification). Full protocol and dataset: our case study on crawl budget for a brand-new site.
Verifying it's really ChatGPT-User
OpenAI publishes dedicated IP ranges for ChatGPT-User — the largest of its three agents, since requests originate from its serving infrastructure. IP verification remains the only proof of authenticity: our analyzer runs it automatically, in your browser, and labels each hit as verified, doubtful or spoofed.
View ChatGPT-User's official IP ranges (JSON)Blocking or allowing ChatGPT-User (robots.txt)
Blocking ChatGPT-User amounts to refusing a ChatGPT user the ability to open your pages through the assistant — like turning away a reader. Unless your content is strictly private, blocking has no upside: these fetches produce linked citations in real conversations.
To block it:
User-agent: ChatGPT-User Disallow: /
To explicitly allow it:
User-agent: ChatGPT-User Allow: /
Frequently asked questions
Is ChatGPT-User a crawler?
Not in the classic sense. It doesn't autonomously explore the web: each request is triggered by a ChatGPT user needing the content of a specific page, right now. OpenAI documents it as a user-triggered agent, distinct from GPTBot (training) and OAI-SearchBot (search index).
Why don't I see this traffic in Google Analytics?
ChatGPT-User doesn't execute JavaScript, so no analytics tag fires. Only your server logs record its visits. This is true of almost every AI agent: measuring AI traffic requires log analysis, not tag-based tools.
How do I measure ChatGPT-User visits on my site?
Drop your access.log files into our analyzer: it isolates ChatGPT-User, verifies each hit's authenticity against official IP ranges and lists the pages fetched — all in your browser, no upload.
Does a ChatGPT-User hit mean I was cited?
Very likely but not guaranteed: the agent fetches the page to build its answer, and ChatGPT generally cites its sources with a link. Even a fetch without a citation is a real read of your content by a user.
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 ChatGPT-User, verifies its authenticity and lists the pages it requests — entirely in your browser, no upload.
Analyze my logs