How to get cited by Claude
Three Claude bots since Anthropic's February 2026 restructure: ClaudeBot (training), Claude-User (runtime fetch on user reference), and Claude-SearchBot(index for web-search mode). Allow Claude-User + Claude-SearchBot in robots.txt for citation eligibility. Block ClaudeBot if you don't want training-data ingestion. Add Organization JSON-LD so Claude can attribute the brand correctly.
Run a free Claude citability audit. Checks all three bots plus the GEO fundamentals.
The three Claude bots, decoded
ClaudeBot — training-data crawler. Used to build the corpus that improves future Claude model versions. Blocking opts out of training but does not affect citation.
Claude-User — runtime crawler. Fires when a Claude user explicitly references a URL mid-conversation (e.g. “summarize this page: example.com”). Blocking opts out of this user-driven citation.
Claude-SearchBot — index crawler for Claude's web-search mode (launched Feb 2026). Builds the retrieval corpus that powers Claude's search-result citations when web search is enabled.
The legacy Claude-Web identifier is deprecated but still honored — Anthropic kept the alias to avoid breaking robots.txt files that block it.
Robots.txt template for Claude (citation, no training)
# Block training-data ingestion
User-agent: ClaudeBot
Disallow: /
# Allow citation crawlers
User-agent: Claude-User
Allow: /
User-agent: Claude-SearchBot
Allow: /If you want both training and citation, simply allow all three. If you want neither, disallow all three. The targeted pattern above is the "cited but not trained on" configuration.
FAQ
What are Claude's crawlers?
Since Anthropic's February 2026 restructure, three distinct Claude bots exist: ClaudeBot (training data crawler — used for model improvement), Claude-User (real-time crawler — fetches a URL when a user explicitly references it mid-conversation), and Claude-SearchBot (index crawler — builds Claude's retrieval corpus for the new web-search mode). Each is a separate User-Agent; treat them independently in robots.txt. The legacy Claude-Web identifier is deprecated but still respected for backward compatibility.
Can I block ClaudeBot but allow citation?
Yes. ClaudeBot is the training-data crawler; Claude-User and Claude-SearchBot handle citation. Block ClaudeBot if you don't want your content used for model improvement, but allow the other two for citation eligibility. The targeted robots.txt pattern: `User-agent: ClaudeBot / Disallow: /` followed by `User-agent: Claude-User / Allow: /` and `User-agent: Claude-SearchBot / Allow: /`.
Does Claude cite during normal conversations?
Claude cites when web search is enabled (Claude Pro / Team / Enterprise users) and when a user-pasted URL is referenced. Without web search active, Claude answers from its training cutoff — citation requires the runtime crawlers to fetch live content. As of mid-2026, the share of Claude conversations that trigger web search is roughly 15-25% (growing), so allowing citation gets you a meaningful fraction of Claude traffic.
How does Claude-SearchBot differ from PerplexityBot or OAI-SearchBot?
Mechanically similar — all three build retrieval indexes for runtime AI lookups. Claude-SearchBot is newer (Feb 2026) and crawls less aggressively than the other two. Coverage is improving fast; the bot list at Seoxpert's geo-citability check is updated every 6 months as Anthropic, OpenAI, and Perplexity refine their crawler architectures.
Does Claude require special schema?
No Claude-specific schema. The general GEO fundamentals apply: Organization JSON-LD on the homepage (so Claude knows which entity you are), Article or HowTo schema on content pages, FAQPage where relevant. Claude reads the same structured data formats as ChatGPT and Perplexity.
Audit your Claude citability
Free first scan. Checks all three Claude bots and the GEO fundamentals.