What Claude tells buyers about you

Fewer people, further along. This is where the careful question gets asked, the one somebody types before spending real money, often on behalf of someone else.

We read: API + web search · checked 2026-08

How people actually use Claude

Longer questions, and several at once. “Compare these three suppliers on lead time, minimum order, and whether they will do a one-off custom job” is a Claude question. The answer is expected to hold up to being read twice.
Frequently at work, on a company account, with the reply pasted into a document that somebody else will act on. The person who asked is often not the person who decides, which means your name has to survive being summarised a second time.
It is reluctant to crown a single winner. Where another engine picks one and moves on, this one tends to lay out several options and say what each is for. That is good news if you are on the list and worse news than it looks if the list is short.

Where its answers come from

Claude · API + web search

The API with web search switched on, said plainly on every result, and the same caveat applies as everywhere else: the API is not the app somebody has open on their laptop. One more thing belongs here, because a page about this engine that left it out would be conspicuous. The same company’s models do work inside this product that has nothing to do with the engine: sorting answers into the four verdicts, and writing the plain-English narrative in your report. That is on the methodology page as well. It gives this engine no advantage in your results: the classifier is never told which engine an answer came from, so it cannot favour one, and the same prompt reads all five.

This is stated on every result you get, not only here. See how the score is calculated.

What we measure, and what counts as good

Every answer Claude gives is graded on the same four-step scale as every other engine, so your columns are comparable. The scale is ordinal and every step carries a word and a mark as well as a shade — nothing here is colour alone.

AbsentYou are not in the answer at all
NamedMentioned, one of several
RecommendedAI tells them to choose you
CitedYour site is the source it quotes
Predates your last change

What to expect on Claude

This engine answers at length and hedges more than the others, so a business can be named twice in three paragraphs and still not be the one being recommended. The verdict tracks what the answer actually urges the reader to do, not how many times your name appears in it.

What moves this engine

In the order worth doing them. These are checks from the scan itself, not general advice. Each one is graded on your site and each one comes with the change written out.

1Let AI systems read your websiteIf ChatGPT and Perplexity cannot fetch your pages when someone asks about you, nothing else on this list matters.about 2 minutes
2Answer your customer questionsEngines lift sentences that answer the question directly. Pages that describe the company instead get skipped in favour of a competitor who simply says the price.about 120 minutes
3Answer more of the questions customers askAnswer-first content is what gets quoted. Prose that buries the answer in paragraph four does not.about 45 minutes
4Help AI understand your servicesStructured data is the most reliable way to state facts about your business that a model will not misread.about 30 minutes

15 checks run on your website in total, and the scan reports every one of them. These are the ones that matter most for this engine.

What we will not tell you

What share of your customers use Claude. Nobody can measure that for your business, ourselves included, and a percentage invented to sound authoritative is worse than no number at all. Your scorecard shows each engine separately so you can weigh them against what you know about your own buyers.
That we can guarantee an answer changes. We can tell you what it says today, what the engines cite instead of you, and which specific things about your site are stopping them. Then we re-scan and show you the before and the after. Anyone promising a position is selling you something they do not control.
That a single scan settles it. These answers move. One reading is a snapshot, which is why the licence includes a re-scan and why monitoring exists as something you can add rather than something we pretend one scan already is.

Common questions

You use this company’s models yourself. Does that bias my score?

It cannot, and the reason is structural rather than a promise. The classifier that turns an answer into a verdict is never given the name of the engine that produced it. The engine is recorded on the row for provenance and is deliberately left out of the prompt. It reads five answers with one set of rules and no way to know which is which. The deterministic half runs first anyway: if your domain is in the engine’s own citations, that is settled before any model is asked.

Hardly any of my customers use this one. Why does it count?

Because of who does. This engine is used disproportionately inside businesses, for the researched, written-up, spend-real-money decision, the kind where one answer is read by four people. If you sell to other businesses, a weak result here matters more than the volume suggests. If you sell to consumers on a Saturday morning, it matters less, and your scorecard shows each engine separately so you can weigh that yourself rather than take our word for it.

What actually changes an answer here?

The same three things as everywhere, in the same order: let the crawlers in, put the facts in readable text on the page, and answer the real questions in words a machine can lift. This engine leans harder than most on pages that answer a question directly, which is why a genuine FAQ (real questions, real answers, not a keyword list) does more here than it does elsewhere.

The other engines

The quickest way to find out is to look at your own result.

Scan my site free

Claude is a trade mark of its owner. MentionMachine is not affiliated with, endorsed by or connected to it, and the mark drawn beside its name on this page is our own. What is described here about how Claude behaves was last checked 2026-08; if something on this page is out of date or wrong, tell us and we will correct it.