How to get cited by Claude

Claude can search the public web in supported experiences and provide citations. That creates an opportunity for publishers, but there is no guaranteed inclusion mechanism. The practical work has three layers: permit the access you want, create a source worth retrieving and prove whether Claude actually mentions or cites it.

Understand Anthropic's three crawler roles

Anthropic's official crawler documentation distinguishes three user agents:

This separation matters. A company may want search visibility and user-requested retrieval while opting out of future training collection. Do not use one blanket rule without understanding the tradeoff.

Audit access beyond robots.txt

An allowed user agent can still fail to retrieve the page. Test:

• status codes and redirect chains
• WAF, CDN and rate-limit behavior
• server-rendered main content
• canonical and noindex directives
• accidental blocks on scripts or data required to render the answer
• geographic or cookie barriers

Use server logs to distinguish a crawl problem from a content-selection problem. If the bot receives a clean 200 response but Claude never cites the page, publishing quality and authority become the next suspects.

Create a page Claude can cite precisely

Web-enabled answers need source material that supports specific statements. Make important facts visible, dated and attributable.

For commercial queries, cover the real decision. A product comparison should explain who each option is for, which capabilities were tested, what was excluded and when the information was verified.

Build a source architecture

One long page cannot be the best source for every claim. Create a small set of canonical destinations:

1. a product page for capabilities and audience
2. a pricing page for current plans and limits
3. a methodology page for measurement definitions
4. focused guides for common buyer problems
5. documentation for implementation facts

Connect them with descriptive internal links. Avoid copying the same paragraphs across every page, because duplicate versions make the canonical source less obvious.

Add independent evidence

Claude may retrieve sources beyond the brand domain. Earn credible third-party coverage through customer stories, integrations, directories, industry research and expert reviews. Correct inaccurate listings and keep official profiles current.

The objective is corroboration, not mention volume. A specific review that explains a tested workflow is more useful than dozens of generic directory entries.

Measure Claude visibility correctly

Build a stable prompt set across discovery, comparison, objections and purchase intent. Record:

• whether web search or research was enabled
• whether the brand appeared
• whether Claude recommended or merely named it
• every citation URL
• competitors and their cited sources
• date, language, market and product surface

Anthropic says its Research experience can conduct multiple searches and return citations. That makes the exact mode part of the test context. A plain answer and a Research answer are not interchangeable observations.

Use the AI share of voice framework to compare competitors and the AI crawler guide to review access across providers. For a concise engine checklist, visit how to appear in Claude.