Get cited by Claude
How to get your site cited in Claude's answers
Claude searches the web, reads pages for users on demand, and speaks MCP — three separate doors. Sites that are legible to machines and honest with evidence walk through all three.
How Claude finds sources
Anthropic operates the same three-crawler pattern as OpenAI: ClaudeBot for training, Claude-SearchBot for the retrieval index behind Claude's web answers, and Claude-User for live fetches when a person's conversation needs your page. Anthropic documents these agents and respects robots.txt.
Claude is notably evidence-first: answers favor sources whose claims are attributable — a page that quotes the statute, names the number, and links the primary source is a materially better retrieval candidate than one that paraphrases.
Claude also connects to tools via MCP (Model Context Protocol). A site that exposes real utilities over MCP — like ours does — can be used by Claude directly, which is a form of presence no ranking system offers.
| Agent | Role | What it means for you |
|---|---|---|
| Claude-SearchBot | Search index | Builds the index behind Claude's web answers. Never block this if you want citations. |
| Claude-User | Live fetch | Fetches your page when a user's Claude conversation needs it. |
| ClaudeBot | Training | Training crawler. Blocking affects future model knowledge, not today's answers. |
The Claude playbook
- 1Verify crawler accessrobots.txt must allow Claude-SearchBot and Claude-User. Our free checker tests ClaudeBot access explicitly, alongside the rest of the AI crawler roster.
- 2Publish llms.txtAnthropic's ecosystem popularized llms.txt. A tight, accurate one — what you are, what you're authoritative on, your key URLs — is cheap and directly useful to Claude. Generate one free with our tool.
- 3Lead with verifiable claimsVerbatim citations, dated facts, named sources. This is the house method we run for every tenant: no claim without a primary source. It is also precisely what evidence-first retrieval rewards.
- 4Structure for machine skimmingQuestion-shaped H2s, a TL;DR up top, FAQPage + Article schema, stable anchors. Claude reads fast; make the skeleton carry the meaning.
- 5Consider an MCP surfaceIf your product has a checkable utility, exposing it via MCP puts you inside the assistant instead of just in its citations — we run ours at /api/mcp and documented the pattern.
What doesn’t work (so you don’t pay for it)
- Prompt-injection tricks — instructions hidden in your pages are treated as data, not commands, and burn trust.
- Thin AI-generated pages at scale — evidence-free volume is exactly what retrieval filters out.
Claude, answered plainly
How do I get listed in Claude's answers?
Allow Claude-SearchBot and Claude-User, publish llms.txt, make every important page question-shaped and citation-dense, and add FAQ/Article schema. Then run a crawl check to confirm nothing is blocked.
What is llms.txt and does Claude use it?
llms.txt is a plain-text map of your site written for AI assistants — what you are, what you cover, your canonical URLs. It emerged from the Anthropic ecosystem and is read by multiple assistants. Ours is at /llms.txt; you can generate yours free.
Does Claude respect robots.txt?
Anthropic documents its crawlers and honors robots.txt directives for them — which means your robots file genuinely controls your Claude presence, in both directions.
More engines: ChatGPT · Perplexity · Gemini · Grok · Copilot · the full GEO guide
Can Claude read your site today?
17 checks in about 60 seconds, no signup — crawler access, structure, schema, and answer-readiness, graded with exact fixes.