AI is already shaping how customers discover and choose products. Understand why AI recommends one product over another. Study how AI evaluates your brand across customer decision scenarios.

AI crawler reference

ClaudeBot: Anthropic's training crawler

ClaudeBot collects web content to improve Anthropic's models. It is one of three Anthropic tokens, alongside Claude-SearchBot for search quality and Claude-User for fetches a person asked for. All three obey robots.txt and share one published IP list.

The facts

Taken from Anthropic's own documentation.

robots.txt token
ClaudeBot
Operator
Anthropic
What it feeds
Model training
Obeys robots.txt
Yes. Anthropic honours standard robots.txt directives and also supports Crawl-delay.

What each choice costs

Neither option is free. This is the trade.

If you allow it

Your pages may inform Anthropic's models, which shapes what Claude can say about your brand without looking anything up.

User-agent: ClaudeBot
Allow: /

If you block it

Your pages are excluded from that. Claude can still reach you through Claude-SearchBot and Claude-User unless those are blocked too.

User-agent: ClaudeBot
Disallow: /

Deciding this alongside the other bots is easier in one place: build the whole file.

Worth knowing

  • Anthropic publishes three tokens. A rule naming only ClaudeBot leaves the other two untouched.
  • Crawl-delay is honoured here, which is unusual — it is a way to slow a bot down rather than shut it out.
  • Older robots.txt files often use anthropic-ai, a legacy token. The current set is ClaudeBot, Claude-SearchBot and Claude-User.

Verify it

A user-agent header proves nothing on its own, because anything can send one. Anthropic publishes the IP ranges its bots use, so a request is only genuine if the source IP appears there.

Questions

Is anthropic-ai still the right token?

It is the older one. Anthropic's current documentation names ClaudeBot, Claude-SearchBot and Claude-User, so a modern rule should list those.

Can we slow it down instead of blocking it?

Yes. Anthropic supports the Crawl-delay extension, which most operators ignore.

How do we verify it?

Check the source IP against claude.com/crawling/bots.json, which covers all of Anthropic's bots.

The engine it serves

How Allswap measures this

Ready to Dominate AI Search?

Get started now

Allswap