How to Get Cited by Claude (Anthropic) in 2026
Claude's web retrieval runs on Brave Search — not Google, not Bing. Research from Profound (2025) found an 86.7% overlap between Brave Search results and Claude's real-time web citations. This makes Claude the most predictable AI engine to optimize for: get indexed in Brave, allow the right crawlers, and structure your content clearly.
Why Brave Search is the key to Claude citations
Anthropic built Claude's web retrieval around Brave Search's independent index. Brave is privacy-focused and builds its own web index rather than licensing from Google or Microsoft — meaning it has completely different coverage than Google and Bing.
Many well-optimized websites have excellent Google coverage but incomplete Brave coverage. Since Claude pulls from Brave, those sites appear in Google searches but are effectively invisible to Claude's web retrieval.
Action item #1: Submit your sitemap to Brave Search at search.brave.com/submit. This is free, takes 2 minutes, and is the single highest-leverage technical action you can take to improve Claude citations.
Anthropic's three-crawler framework
Like OpenAI and Perplexity, Anthropic uses separate bots for different purposes:
| Crawler | Purpose | Affects Claude citations? |
|---|---|---|
| ClaudeBot | Training data | No |
| Claude-SearchBot | Web search during Claude queries | Yes — primary |
| Claude-User | Real-time browsing | Yes — secondary |
| anthropic-ai | Legacy string (older) | Keep for compat |
| Claude-Web | Legacy string (older) | Keep for compat |
Your robots.txt needs explicit Allow entries for all three active bots. A common mistake: the site allows ClaudeBot (training) but blocks or has no rule for Claude-SearchBot and Claude-User, which are the two that actually affect citation visibility.
What Claude weights in citation decisions
Beyond Brave Search indexing, Claude's citation selection favors:
Entity clarity and brand disambiguation
Claude is trained to cite sources it can identify with confidence. Pages that have clear entity signals — Organization schema with a name, URL, description, and sameAs links — score higher for citation confidence. Ambiguous brands with no schema are cited less frequently, and when cited, more often with incorrect descriptions.
Factual, sourced content with specific claims
Claude's citation model favors content that makes specific, verifiable claims. Vague marketing copy ("the best solution for your needs") gets passed over. Specific data ("reduces audit time by 73%", "covers 35 AI crawlers") gets cited. Write in facts, not promises.
llms.txt — Anthropic explicitly supports the spec
Anthropic is one of the primary organizations behind the llmstxt.org specification. Claude actively reads and processes llms.txt files during web retrieval sessions. A properly formatted llms.txt is the most direct signal you can send to Claude about your brand and content.
Clean, fast HTML pages
Claude's browsing agent parses page content in real time. JavaScript-heavy SPAs that render content client-side are harder to parse. Static or server-rendered HTML that loads in under 2 seconds gets more complete extraction.
The complete Claude citation checklist
search.brave.com/submit — include all key content pages
Plus ClaudeBot, anthropic-ai, Claude-Web for completeness
Blockquote summary + key page links per llmstxt.org spec
LinkedIn, Crunchbase, ProductHunt, GitHub — entity graph breadcrumbs
Replace vague benefit language with specific, sourced claims
Mirror the exact question phrasing users ask about your category
Is Claude finding your site?
Neurobird checks your Brave Search indexability, Anthropic crawler configuration, llms.txt validity, and entity schema — free in 30 seconds.
Run free audit →