How to Get Recommended by Claude

Claude is the careful one

Every AI tool has a personality, whether its makers intended that or not, and Claude’s comes across as more cautious than most about stating things it can’t back up. Ask it to recommend a specific business and, more often than the others, it’ll hedge, qualify its answer, suggest you verify details yourself, or decline to name just one option if the evidence in front of it feels thin or overly promotional. That’s by design, Anthropic has built Claude to lean towards being useful without overstating its confidence. For a business owner, that caution cuts both ways. It means Claude is harder to win over with polished marketing language alone, phrases like ‘award-winning’ or ‘industry leading’ tend to read as unverifiable claims rather than evidence, and Claude is more likely to quietly discount them. But it also means that once you’ve given Claude something genuinely checkable, it’s often more willing to state it plainly, because the caution is about unverifiable claims, not about being negative toward businesses generally.

What actually convinces a cautious model

Specific, boring, checkable facts do more work with Claude than anywhere else on this list. Opening hours. A real price range. How long you’ve been trading. The actual qualifications your team holds. The postcode areas you cover. None of that is exciting copy, but it’s exactly the kind of detail a cautious model can treat as fact rather than opinion. Consistency matters just as much, if your ‘12 years trading’ claim on your website doesn’t match what a directory listing says, or your service area on one page contradicts another, that’s the kind of contradiction that pushes a careful model further toward hedging rather than committing to a recommendation. The businesses that come across best to Claude tend to be the ones that write about themselves the way they’d fill in an official form, plainly, accurately, without embellishment, then repeat those same facts identically everywhere else they appear online.

Where Claude shows up beyond the chat window

It’s worth remembering Claude isn’t only the chat interface at claude.ai. Anthropic’s models sit behind a growing number of workplace tools, research assistants and customer-facing products that businesses build using their API, which means someone using a piece of software to shortlist suppliers, research a service, or draft a recommendation might be running on Claude underneath without ever seeing the Anthropic name at all. That matters for procurement-style searches in particular, someone at another company quietly researching who to hire, rather than a consumer chatting casually. Being ‘Claude-legible’, clear, factual, consistent, isn’t just about pleasing one chatbot, it’s increasingly about being legible to a layer of software sitting underneath tools you’ll never see the inside of.

How to write for a model that double-checks everything

Go through your own website and flag every unsubstantiated superlative, ‘leading’, ‘best’, ‘number one’, and either replace it with the real fact behind the claim or cut it. If you genuinely are the longest-established firm in your trade locally, say since when, and make sure that date matches your Companies House record and your ‘about us’ page word for word. If a claim has a source, a certification body, an award you actually won, an accreditation, link to it or name it properly rather than just asserting it. Then check your details match across every listing you appear in, directories, socials, review sites, because a cautious model treats matching independent sources as far stronger evidence than one confident sentence on your own homepage.