Does Claude cite sources, and can you be one?

The two answers, again

Like every major assistant, Claude answers in two modes, and they reward different things.

From training. Without a live search, Claude draws on what it learned from reading the public web: which businesses are associated with which work, which explanations are standard, which names recur with which credibility. Your presence in training-based answers is your accumulated public footprint; it moves slowly, in both directions, which makes it worth building and hard for a competitor to take.

From live search. When the question needs current information, Claude searches, reads a set of pages and answers with citations linked to what it used. This is the winnable surface: it responds to the same underlying qualities the other engines reward, findability, extractability and entity clarity, with an emphasis worth understanding on its own terms.

On the plumbing, Anthropic documents three separate crawlers, each controllable individually in robots.txt, and its documentation states plainly what blocking each one costs. ClaudeBot gathers material for training; Claude-SearchBot indexes content for Claude's search results; Claude-User fetches a page when a user's question sends Claude to it directly. The practical point for visibility: a blanket AI block left over from 2023 typically blocks all three, and it is the second two, Claude-SearchBot and Claude-User, that carry your presence in Claude's answers. Anthropic's own wording is that blocking them may reduce your site's visibility in user search results. Check what your robots.txt actually says; the fix takes minutes.

What Claude's citation behaviour rewards

Checking Claude's answers by hand across client categories, three tendencies stand out, and each is actionable.

A leaning toward primary and authoritative sources. Claude is noticeably disposed to cite the organisation itself for claims about the organisation, official documentation for technical claims, and recognised references for factual ones. For a business this is good news with a condition attached: your own website is the natural citation for who you are, what you do and what you have done, provided it states those things checkably. A claims page with numbers, named work and dates gives Claude something citable; a wall of adjectives does not. Our case-study pages exist in the structure they do partly for this reason: verifiable specifics, one per claim, on a stable address.

Care with uncertainty. Claude is built to be cautious about asserting what it cannot support, which means ambiguous entities lose twice: not only is a blurry business less likely to be named, the caution compounds it. The entity work, one consistent name, address, service set and story across your site, your profiles and third-party mentions, backed by structured data that says the same thing, is covered fully on our GEO page and matters here as much as anywhere.

Substance over polish. The pages we see cited are dense with answerable material: definitions, steps, numbers, dates, named examples. Length is not the variable; liftability is. A 500-word page that answers cleanly beats a 3,000-word page that circles.

Two honest limits to this picture. Anthropic publishes less operational detail about its answer-composition layer than Google does about its features, so where the mechanics are not documented we describe observed behaviour and label it as such. And Claude's product surface changes; this page carries a last-reviewed date and gets updated on this address, per the standard this whole series runs on.

Why Claude is worth optimising for at all

A fair question, since Claude's search volume is smaller than Google's. The answer is who asks and what they ask. Claude's usage skews toward considered, professional queries: comparing suppliers properly, understanding a technical purchase, doing the due diligence a referral triggers. Those are exactly the moments a growth-stage business is chosen or passed over. A citation in front of one genuinely evaluating buyer is worth more than impressions in front of a thousand browsers, and the work that earns it is the same authority work the rest of your search visibility runs on. Nothing here is a separate invoice; it is the same AI search discipline pointed at one more surface.

What we do about it, and what you can do today

We make the checkable version of your business easy to find and easy to trust: claims pages with verifiable specifics, entity consistency across every surface that mentions you, structured data that corroborates the words, and authority earned from sources engines already respect. Then we ask Claude the questions your buyers ask and record what it actually says, by hand, because that is the only measurement that reflects reality. The AI-search visibility check covers Claude alongside the other engines and is free.