Claude is Anthropic's AI assistant, widely used in enterprise settings for research, analysis, and content synthesis. Claude is trained on a large corpus of web content and, in its tool-use and web-search modes, actively fetches and cites web pages. Getting cited by Claude is particularly valuable for B2B companies targeting enterprise buyers.
Claude's citation behavior comes from two sources: training data (crawled by ClaudeBot / anthropic-ai crawler before training cutoff) and real-time web access (in Claude Pro and enterprise versions with web search enabled). Claude's training emphasizes accuracy and source quality — it actively avoids citing content that appears promotional, superficial, or self-contradictory. Clear, factual, well-structured content performs best.
ClaudeBotAnthropic uses two bot identifiers: ClaudeBot and anthropic-ai. Both must be allowed for complete coverage.
Anthropic actively supports the llms.txt standard. Claude's API-based tooling explicitly fetches /llms.txt. A well-formed file gives Claude a curated orientation to your company before it reads individual pages.
Claude's training penalizes content where it cannot confidently identify who the author is, what company is being described, or how a product name maps to a company. Be explicit: "Ansly, made by [Company], is an AEO audit platform" — not just "the platform".
Claude is frequently used for technical evaluation queries. Sites with detailed, accurate documentation (how it works, integrations, API reference) are cited significantly more than marketing-only sites.
The entities.txt format is particularly useful for Claude, which builds entity maps of the content it reads. A well-formed entities.txt reduces ambiguity and cross-references your brand with related entities Claude already knows about.
Claude's training data selection weighs source quality. First-person research, original data, case studies, and specific statistics are more citation-worthy than general advice that could be found anywhere.
Claude's web access mode fetches raw HTML. JavaScript-rendered content is not visible. Use SSR or SSG.
Only blocking anthropic-ai but not ClaudeBot (or vice versa)
Anthropic uses both bot names. Check robots.txt for both "User-agent: ClaudeBot" and "User-agent: anthropic-ai" and ensure neither has Disallow: /.
Marketing copy that avoids making specific claims
Claude deprioritizes vague promotional language. Replace "industry-leading platform" with "platform that processes X audits per month" or "trusted by 500+ companies in [category]".
No documentation or technical content
Add a /docs or /how-it-works section with technical detail. Even a single well-written "How it works" page dramatically improves Claude citation rates for evaluative queries.
Claude Pro and enterprise plans include web search capability. In this mode, Claude fetches real-time web content when answering queries, using DuckDuckGo or Bing as a retrieval layer. Even without web browsing, Claude's training data (collected by ClaudeBot) includes web content up to its training cutoff.
Claude hallucination about specific companies usually occurs because: (1) your company is small enough that training data is sparse, (2) your company name is ambiguous (shared with another entity), or (3) your web content is contradictory or vague. The fix: create an unambiguous Organization schema, a detailed llms.txt, and an entities.txt that cross-references your brand with known entities.
ansly audits your site for all the signals Claude uses to cite content. Get your free AEO score in 60 seconds.
Audit my site free →