Findings · 2026-11-13 · By VSNARY | Emmanuel Orta · 1 view
Claude-User uses two published address lists. We checked one
Anthropic publishes a crawler address list and a separate list of outbound addresses. Claude-User can arrive from either. We compared against the first only, and filed real requests as failed.
Until 13 November 2026 CrawlCheck checked requests named Claude-User against one list only, Anthropic's crawler file. Anthropic also publishes outbound address ranges, 160.79.104.0/21 and 153.61.208.0/20, and Claude-User fetches arrive from those too. Requests from the second list were counted as failing verification. The check now accepts both lists for Claude-User. 602 earlier rows remain filed as failed and have not been reclassified.
Claude-User is the fetcher Anthropic uses when a person asks Claude to open a page. We verify a crawler by comparing the request's address with what its operator publishes. For Claude-User we compared against one list when there are two.
The two lists #
| List | Published as | We checked it |
|---|---|---|
| Crawler addresses for ClaudeBot, Claude-SearchBot and Claude-User | A JSON file on claude.com | Yes |
| Outbound address ranges: 160.79.104.0/21 and 153.61.208.0/20 | A page in Anthropic's API documentation | No, until 13 November 2026 |
How we found it #
We asked Claude to fetch pages on this site and watched the requests arrive. They carried the Claude-User name and came from 160.79.106.14 and 153.61.208.9. Both addresses sit inside the outbound ranges Anthropic publishes. Our record marked them as failing verification.
What changed #
Claude-User is now accepted from either list. The change is narrow on purpose: a request named ClaudeBot from the outbound ranges still fails, because Anthropic lists ClaudeBot's addresses in the crawler file only. An address outside both lists still fails, and so does an address Anthropic has retired.
What is still wrong #
Before the fix, 602 requests named Claude-User were filed as failing. Some of those were real. We have not gone back and reclassified them, so the historical failed count for Claude-User is too high. We have also not yet confirmed the fix against a visit we did not trigger ourselves.
If you check crawlers yourself #
The lesson is general: an operator can publish more than one list, and on-demand fetchers are the likeliest to use a different one from the index crawler. Before you block a fetcher for failing an address check, look for a second list in the operator's documentation. Our mirror of operator lists covers the crawler files; it does not replace reading the operator's pages.
And remember that what each operator says about robots.txt differs for this kind of fetcher too. Anthropic says Claude-User follows it.
CrawlCheck’s dataset when this page loaded: 570,959 sites measured.
Share this finding:LinkedInXEmailFollow Emmanuel on LinkedIn
Questions this post answers
What IP addresses does Claude-User come from?
Anthropic publishes a crawler address file on claude.com and, separately, outbound ranges 160.79.104.0/21 and 153.61.208.0/20 in its API documentation. Claude-User requests can arrive from either.
Why did CrawlCheck mark real Claude-User requests as failed?
It compared them with the crawler file only. Requests from the outbound ranges were not on that list, so they were counted as failing verification until 13 November 2026.
Does ClaudeBot use the outbound ranges too?
We do not accept it from them. Anthropic lists ClaudeBot's addresses in its crawler file, so a ClaudeBot request from the outbound ranges still fails our check.
Were the old Claude-User counts corrected?
No. 602 earlier requests remain filed as failed. Some were real, so the historical failed count for Claude-User is overstated.
Related findings
The main product
Found this on your own site? We fix it for $100.
Scan free to see where you stand. The fix is one site, every finding implemented and re-measured, with a sealed before and after.
Free browser extension
Check this on any site you open, without leaving the page.
CrawlCheck for Chrome grades every site you visit A–F and shows, from your own browser, what search engines and AI assistants are actually served. Free, no account.
Write for the library
Measured something yourself? Submit an article.
Send your finding, the data and the method. A person reads every submission and replies by email; accepted articles are published under your name with a link to your study. Free to submit. Articles are accepted on merit; no one can pay to have one published.
How it works · a reply within about a week
Spotted a mistake or a bug on this page? Report an error. Confirmed errors are fixed and logged on the corrections page.
Comments
Comments are read before they appear. Nothing is published automatically, and no account is needed. An AI agent with a published signing key can post without the queue: it signs the request and the signature is checked against its own domain (how). Those appear at once, marked verified.
Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.