Claude-User: what this user agent does and what blocking it means

99Visibility Research · Updated October 21, 2026 · 4 min read

Short answer

Claude-User is Anthropic's user agent for web fetches that a person triggers by asking Claude to read or check a specific page. It is separate from ClaudeBot, which collects training data, and Claude-SearchBot, which supports Claude's search quality. If you block Claude-User, Claude may be unable to read your page when someone asks about it, and its answer will not draw on your own words.

Key takeaways

Claude-User fetches a page because a person asked Claude to, not to build a training set or a search index.
Anthropic runs three separate agents (ClaudeBot, Claude-SearchBot, Claude-User), and each one needs its own robots.txt decision.
A firewall or CDN can turn Claude-User away even when robots.txt allows it, so test both layers.
Allowing Claude-User makes your page readable on request. It does not guarantee that Claude cites or recommends you.

Definition

Claude-User is the user agent Anthropic uses when Claude fetches a web page on behalf of a person in a conversation. Anthropic's crawler documentation (updated 2026-04-07) lists it alongside two other agents with different jobs:

ClaudeBot collects content that may be used for training.
Claude-SearchBot supports the quality of Claude's search results.
Claude-User handles fetches a user triggers, such as "read this URL and summarize it".

That makes Claude-User a user-triggered fetcher, the same category as OpenAI's ChatGPT-User and Perplexity's Perplexity-User. For a full breakdown of all three Anthropic agents and how to configure them, see ClaudeBot, Claude-SearchBot and Claude-User explained.

How a Claude-User fetch works

A person pastes your URL into Claude, or asks a question that needs a specific page. Claude requests that page in real time, identifying itself as Claude-User, and uses what comes back in its answer.

Two gates decide whether that request succeeds:

robots.txt. Rules are per bot. A rule for ClaudeBot says nothing about Claude-User. If you want to address Claude-User directly, name it:

User-agent: Claude-User
Allow: /

The network layer. A firewall, CDN or bot-management rule can return a 403 or a challenge page to an agent that robots.txt allows. Anthropic also warns that IP blocking impedes its ability to read your robots.txt file at all.

The content itself is a third issue. Independent testing as of June 2026 found that ClaudeBot and Claude-SearchBot do not execute JavaScript. That testing did not cover Claude-User, so the safe move is to put your key facts (prices, plans, a one-line description) in the HTML before any script runs.

Why it matters for AI visibility

Claude-User traffic is the moment a real person is asking Claude about you. Say a buyer pastes your pricing page into Claude and asks which plan fits a team of ten. If Claude-User is blocked, Claude cannot read that page. It falls back on what the model already knows, which is frozen at a knowledge cutoff, or on other pages it can reach.

That is where wrong prices and outdated product names come from. You cannot control what the model learned in training. You can control whether your current page is readable when someone asks.

The next move is simple: check your robots.txt for a Claude-User rule or a blanket block, then request a page as Claude-User to see what your firewall returns. The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away. To confirm that hits in your logs are genuine, use the free AI crawler verification method.

Common misconceptions

"Blocking ClaudeBot blocks Claude." It does not. ClaudeBot governs training collection. Claude-User and Claude-SearchBot are separate agents. See whether blocking ClaudeBot removes you from Claude's answers.
"robots.txt is the whole story." It is necessary, not sufficient. A CDN default or WAF rule can block the fetch silently.
"Every vendor treats user-triggered fetchers the same way." They do not. OpenAI says robots.txt rules "may not apply" to ChatGPT-User, and Perplexity says Perplexity-User "generally ignores" them. Read Anthropic's own page for how it handles Claude-User, because these policies change.
"Allowing Claude-User gets me cited." Allowing it makes your page fetchable on request. What Claude says about the page is decided by the model, not by you.

Frequently asked questions

Is Claude-User a crawler?

Not in the usual sense. A crawler like ClaudeBot moves through the web on its own schedule to collect pages. Claude-User fetches a page only because a person in a Claude conversation asked for it. It behaves more like a browser acting for one user than like an indexing bot, which is why it sits in the user-triggered fetcher category.

Should I block Claude-User?

Only if you do not want Claude reading your pages when people ask about them. For most businesses, those requests come from prospects, customers and journalists checking facts. Blocking them means Claude answers without your current page. If your concern is training use, ClaudeBot is the agent that governs that, and you can block it separately while leaving Claude-User allowed.

How do I know a Claude-User request is real?

A user-agent string is easy to fake, so the name alone proves nothing. Check the request against what Anthropic publishes about its agents and verify the source before trusting or blocking it. Bot-management tools such as Cloudflare classify verified bots using signals like published IP lists, stable user agents or reverse DNS. Treat unverified traffic claiming to be Claude-User with caution.

Does allowing Claude-User help me show up in Claude search results?

Not directly. Claude-User covers fetches a person requests, not the background work behind Claude's search. Anthropic lists Claude-SearchBot as the agent tied to search quality, so check that rule too. Allowing either agent only removes a block. Whether Claude retrieves, uses or cites your page for a given question is decided by Anthropic's systems.

Sources

This page is AI-assisted. It was written from a verified evidence base and checked against our editorial policy. Engine behaviour changes often; facts are as of October 21, 2026.