Claude-User: when a human reads your page through Claude
Claude-User is the agent that visits a web page when a Claude user needs it in their conversation: reading an article, checking a figure, summarizing a document. It is neither an index nor a training crawl — every Claude-User hit corresponds to a real human consultation of your content, performed by the assistant on its user's behalf.
Technical profile
- Operator
- Anthropic
- Role
- Live browsing (user-triggered)
- robots.txt token
Claude-User- User-agent (example)
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Claude-User/1.0; +Claude-User@anthropic.com)- Respects robots.txt
- Yes (documented by Anthropic)
- Official IP ranges
- Not published by the operator (IP verification impossible)
- Official documentation
- Anthropic crawling documentation
The most direct GEO indicator
Claude-User volume is proportional to the actual use of your content in Claude conversations. Unlike training crawls (broad and impersonal), these requests express demand: someone, somewhere, needed your page right now. Tracking them page by page is one of the few concrete ways to measure your presence in the Claude ecosystem.
This traffic completely escapes JavaScript tag-based analytics: the agent doesn't execute scripts. Your server logs are the only source that records it — a textbook example of the value of log analysis in the AI assistant era.
Reading Claude-User's response codes
Watch the errors served to Claude-User: a 404 returned to this agent is a user who couldn't read your content — often through an old link or a stale URL cited somewhere. Keeping clean 301 redirects on your old URLs therefore directly benefits your visibility in Claude.
Verifying it's really Claude-User
Like other Anthropic agents, Claude-User has no published official IP ranges: a hit's authenticity cannot be formally proven. Our analyzer labels these hits as “doubtful” out of honesty, and shows you the pages fetched and response codes served — the signals that actually matter.
Blocking or allowing Claude-User (robots.txt)
Blocking Claude-User amounts to denying Claude users the ability to read your pages — with nothing gained, since this agent serves neither training nor an index. For a site pursuing visibility, it's the last agent you'd want to block.
To block it:
User-agent: Claude-User Disallow: /
To explicitly allow it:
User-agent: Claude-User Allow: /
Frequently asked questions
What does a Claude-User hit in my logs mean?
That a Claude user needed this page in their conversation: the assistant fetched it live to read, verify or cite your content. It's a real human read through AI — the GEO equivalent of an organic visit.
Why doesn't Claude-User appear in Google Analytics?
Because it doesn't execute JavaScript: no analytics tag fires. Only server logs record these visits. To measure your traffic from AI assistants, log analysis is the only reliable instrument.
Are 404s served to Claude-User a problem?
It's a signal worth fixing: every 404 is a Claude user who couldn't reach your content, often through a stale URL. Make sure your old URLs 301-redirect to the new ones — the assistant follows redirects and the read succeeds.
How do I track Claude-User on my site?
Drop your access.log files into our analyzer: it isolates Claude-User, shows pages fetched, frequency and response codes — in your browser, without uploading your logs.
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 Claude-User, verifies its authenticity and lists the pages it requests — entirely in your browser, no upload.
Analyze my logs