Claude SEO: how to get your brand recommended by Claude

With Google you get Search Console. With ChatGPT you get Bing Webmaster Tools sitting underneath it. With most local platforms you can claim a listing and fill in the fields yourself.

Claude gives you none of that.

There is no Claude Business Profile. No submission form. No advertising product. No verification process. Anthropic has built no surface through which a business can tell Claude who it is.

Which leaves exactly one lever: what the rest of the internet says about you, and how clearly it says it.

That sounds like a disadvantage. It is the opposite, for two reasons.

It cannot be bought. No competitor can outspend you into Claude’s answers. The playing field is unusually level.

It compounds into everything else. The work that makes Claude describe you accurately, consistent entity signals, third-party corroboration, clear public positioning, is the same work that improves every other AI platform. Claude is the strictest examiner. Pass it and the rest tend to follow.

Claude also matters more than its raw user numbers suggest. It is heavily used by developers, consultants, researchers, lawyers and analysts: people who are frequently the ones producing the shortlist, writing the recommendation or advising the buyer. Being absent from Claude means being absent from a disproportionately senior and decision-making audience.

Claude answers from two distinct layers, and the difference between them changes everything about what you can influence.

The practical consequence:

If Claude answers a question about your industry from training knowledge alone and does not name you, that is not a page-level problem. No amount of on-page optimization fixes it quickly. It means your brand was not a sufficiently well-established entity across the web when the model was trained.

If Claude searches and still does not surface you, that is usually fixable, and often fast. Crawler access, page structure, or simply nothing existing that answers the question clearly.

Diagnosing which of the two you are dealing with is the first thing we do, and it is the step almost everyone skips.

The three Claude crawlers, and why most sites get this wrong

Anthropic runs three separate crawlers with three separate jobs. They can each be allowed or blocked independently in robots.txt, and blocking one does not block the others.

Three details worth knowing, because they catch people out:

All three honour robots.txt, including the user-initiated one. That is a meaningful difference from OpenAI and Perplexity, both of which warn that robots.txt may not apply to their user-triggered fetchers. With Claude, a robots.txt block genuinely removes you.

Anthropic does not publish IP ranges, and says IP blocking is unreliable because its bots use public cloud provider addresses. Blocking those ranges can stop the bot reading your robots.txt at all. Verify by user agent string instead.

Rules must be set per subdomain. A block on the root domain does not cover shop.yourdomain.com or blog.yourdomain.com.

The mistake we find on roughly half the sites we audit

Most businesses that block Claude never intended to.

It happens one of three ways. Someone read an article about AI training and added a blanket Disallow for ClaudeBot, not realizing the search and user bots are separate. Or a developer enabled an aggressive bot-protection mode on Cloudflare or a similar CDN, which silently blocks AI crawlers regardless of what robots.txt says. Or a WordPress security plugin did it by default and nobody looked.

The result is identical in all three cases: Claude cannot read the site, so Claude cannot recommend the business. Nothing in any analytics dashboard reports this. It is invisible until somebody checks the server logs.

A deliberate choice is different. Publishers with a genuine reason to keep their archive out of training datasets can block ClaudeBot while leaving Claude-SearchBot and Claude-User open. That keeps you visible in answers while opting out of training. It is a perfectly reasonable position, and almost nobody realizes it is available. The full list is in our which AI crawlers to allow.

Here is where most articles on this subject start inventing things.

Anthropic publishes no ranking factors, no scoring model and no weighting for how Claude selects which businesses to mention. There is no equivalent of the Search Quality Rater Guidelines. Any agency presenting you with “the seven Claude ranking factors” has made them up.

So we separate what we know into three honest categories.

What our testing consistently shows

We run prompt sets against Claude for every client and every audit. A few patterns hold up reliably enough to act on:

Claude hedges more than any other model. Ask most assistants for the best supplier in a city and you get three names. Claude will often describe what to look for instead. Getting named at all is a higher bar, which is precisely why it is worth more when it happens.

Description consistency outperforms description quality. A business described in the same words across fifteen sources is named more often than a business described brilliantly but differently on each. Claude appears to be looking for corroboration, not eloquence.

Structured reference sources punch above their weight. Wikipedia, Wikidata and established directories carry more influence than their traffic would suggest. Entity presence on those sources is one of the highest-leverage moves available.

Specificity beats superlatives. “Award-winning digital agency” tells a model nothing it can repeat. “SEO agency in Vancouver specializing in AI search visibility for law firms and dental clinics” gives it something usable. Claude repeats what it can verify, not what you claim.

Answers vary between runs. The same prompt can return different businesses. That variance is itself diagnostic: wide swings usually mean the model is uncertain about your category, which is an opportunity.

What we will not tell you

We will not give you a score for how Claude ranks you against a published algorithm, because no such algorithm has been published.

We will tell you how often Claude names you across a defined prompt set, who it names instead, which sources it pulls from, and whether that number moves after we work on it. That is measurable, repeatable and honest.

It is also considerably more useful than a made-up score.

Two of these matter more than the rest.

Citation sources is the one that turns strategy into a task list. Once you know which twelve domains Claude pulls from when answering questions in your category, you stop guessing. Getting onto those pages becomes the work.

Answer variance is the one nobody else reports. A business named in 8 out of 20 runs is in a very different position from one named in 8 out of 20 consistently every week. The first is on the edge of recognition; the second is reliably mid-pack. The full method is in measuring visibility across repeat runs.

Claude is one platform among six, and optimizing for it in isolation would be a mistake.

Research analyzing millions of AI citations found that around 91% of cited URLs appear in only one model. Strong performance on one platform tells you very little about the others.

Which is why we never sell single-platform optimization. Claude work sits inside our wider generative engine optimization program, alongside ChatGPT, Gemini, Perplexity, Copilot and Grok. Underneath it all sit the same foundations as our SEO agency work: technical health and content strategy.

The useful thing about Claude specifically: because it is the strictest and least gameable of the six, it works well as the benchmark. If Claude describes your business accurately and recommends you confidently, your entity foundations are genuinely sound.

Everything else gets easier from there.