AI crawlers · Anthropic · Training only · Not scored

ClaudeBot

Collects web content that could contribute to training Claude's models. Anthropic says its bots honour robots.txt and support Crawl-delay.

Where this comes from

  • Documented by the vendor. Source: support.claude.com
  • Honours robots.txt. Published addresses: claude.com/crawling/bots.json.
  • Listed, not scored. It cannot keep a site out of today's answers, so the check shows a block without deducting for it.

In the top 1,000 websites

  • 18.3% block it in robots.txt (104 of the 569 sites with a robots.txt), as of 2026-10-12.
  • 14.8% refuse it at the server, answering its user agent with a 401, 403, 429 or 503 while a browser gets the page.
  • 10.9% allow it in robots.txt and refuse it anyway (60 sites): a firewall or CDN rule the site owner may not know about. The reverse also happens: 49 sites block it in robots.txt and still serve it.
  • Method, limits and the full data: the October 2026 study of the top 1,000 websites.

What blocking it costs

Nothing in today's answers, and nothing in the readiness score: ClaudeBot only gathers training data, so blocking it is a policy choice about future models, not about being found. The check lists the block without deducting for it.

The robots.txt lines

To let it in (a group that names the crawler replaces any wildcard rules for it, so this also lifts a User-agent: * block):

User-agent: ClaudeBot
Allow: /

To keep it out:

User-agent: ClaudeBot
Disallow: /

A rule in robots.txt is a request. Anthropic says this crawler honours it. The robots.txt generator writes a whole file with the same split: block the training crawlers, keep the search and answer crawlers.

Also from Anthropic

Does ClaudeBot get through on your site?

Run the free check: it reads your robots.txt against all 51 crawlers and tells you which ones can read you, and what to change.