How to get cited by Claude
Claude, Anthropic's assistant, recommends brands from two sources: patterns learned during training, and — in surfaces where it has web access — live pages it can fetch and quote. You improve your odds by being described consistently across many independent sources, keeping your site crawlable by Anthropic's crawler, and structuring pages so the answer is clear and self-contained. As with every engine, citations are earned, never bought, and the only proof is re-testing the same prompts.
How does Claude decide what to recommend?
When Claude answers from training alone, it relies on what it learned about your category from a broad sweep of the web. Brands that are described clearly and mentioned consistently across many independent sources are the ones it can recall and recommend with confidence. Thin or contradictory coverage produces vague or absent recommendations.
In product surfaces where Claude can search or read the web, it can fetch current pages and ground its answer in them. There, the same fundamentals that help every retrieval engine apply: crawlable, fresh, extractable pages.
Make your site readable to Anthropic's crawler
If you want to be part of what Claude learns and what it can fetch, do not block Anthropic's crawler, and make sure your content is present in raw HTML rather than rendered only in the browser.
- Allow ClaudeBot in robots.txt and at your CDN.
- Serve content server-side so it is visible without running scripts.
- Avoid putting the substance behind logins or interactive widgets.
Write for clarity and extraction
Claude rewards clear, well-organized writing. Lead with a direct answer, use headings that mirror real questions, keep claims specific and sourced, and add Article and FAQPage structured data. A page that states plainly what you are, who you are for, and how you compare gives the model unambiguous material to draw on.
Build broad, consistent corroboration
Because training-based recall depends on how the wider web describes you, the highest-leverage long-term work is earning consistent mentions across reviews, roundups, documentation, and editorial coverage. The goal is for independent sources to describe your brand and category position the same way your own site does.
Track and prove movement
Run your category and comparison prompts in Claude, record who is named, ship a fix, and re-test. Anthropic ships model and product updates over time, so treat this as ongoing measurement rather than a one-time push — and never assume a single good answer is permanent.
Frequently asked questions
It depends on the surface. Some Claude products can search or read the web and ground answers in live pages; others answer from training alone. Optimizing for both crawlability and broad corroboration covers either case.
ClaudeBot is Anthropic's web crawler. Allowing it lets your content be fetched and considered; blocking it removes you from that path.
No. Organic recommendations are earned through how clearly and consistently the web describes your brand, not through payment.
Where Claude reads the live web, a fresh crawlable page can be picked up relatively quickly. Training-based recall changes on the slower timescale of model updates.
Usually because that competitor is described more clearly and consistently across the sources Claude learned from. Closing that description-and-corroboration gap is the fix.