Mentioned vs cited: the two numbers people confuse

A founder checks an AI assistant before a buyer call and sees the product named in the answer, but the company site is absent from the sources. She needs to know whether that is a visibility win or a warning sign.

A mention means the engine named your product. A citation means it linked to a page as evidence. We can only measure where citations land, not how often a product is named without one.

Are you tracking AI mentions and citations separately?

Start with the distinction rather than a combined visibility score. If an assistant names your product while citing another company's comparison page, you have recognition without ownership of the source. If it cites your page without naming the product prominently, you have a source that may be useful but weaker evidence that the product is being selected in the answer. That is the practical difference between a mention and a citation.

The distinction changes the work. A mention problem points you towards the pages and publishers that establish what the product is. A citation problem points you towards pages that answer buyer questions clearly enough for an engine to use them. Combining both into one number hides the route.

Why can your product be named while another site gets cited?

An engine can learn about a product from a third party and use that page as its source. A roundup may name several products, a competitor page may compare them, or a community thread may describe a real use case. The product can therefore enter the answer through surrounding evidence while its own site never becomes the linked source. That is why a named product and a linked page can belong to different publishers.

The citation mix gives this pattern useful context. According to the weekly research on who owns AI answers, listicles and roundups account for 68.3% of recorded citations, with 44,164 citations, while competitor product pages account for 24.1%, with 15,575 citations. Together, those page types take 92.4% of citations in the sample, so the pages shaping an answer often sit outside the product's own domain.

That does not make your own site irrelevant. It means the route into an answer can be broader than your domain, with other publishers supplying the comparison, description or use case that an engine chooses to cite. Study those pages for the buyer questions they serve, the facts they repeat and the language they use to position products.

What does the difference look like in real numbers?

Take the full sample and walk through it line by line. CiteGraph scan data records 64,667 citations across 9,791 AI answers. Those answers come from 77 scans of 44 sites in 48 categories, and the scans asked 507 distinct buyer questions. These figures describe a measured sample, not every industry, engine or prompt, and we do not know what happens outside it.

The next line is the source set: engines cited 9,192 distinct domains and 21,465 distinct pages. Listicles and roundups supplied 44,164 citations, or 68.3% of the total, while competitor product pages supplied 15,575, or 24.1%. The scanned product's own site supplied 1,107 citations, or 1.7%, while community threads supplied 1,930 citations, social posts 538, documentation 509, and review platforms 489, according to the same research data. The shape is clear: linked evidence usually came from other page types, even when a product may have appeared in the answer.

That is the useful comparison. A product can be present in an answer while a roundup or competitor page receives the citation, because the engine has chosen that page to support the claim. A low own-site share therefore tells you where to inspect first, rather than whether the product was recognised at all.

Which sources should you inspect when your own site rarely appears?

Look first at the pages that already carry the category. In the sample, listicles and roundups are the largest source group, followed by competitor product pages. Those are not interchangeable. A roundup may shape the shortlist, while a competitor page may frame the comparison and describe where your product fits.

The table below is a diagnostic frame for your own scan results, not a distribution observed in the CiteGraph sample. It helps you keep the answer outcome and the source outcome separate while you inspect individual questions.

Use the table as a reading aid, not as a promise of causation. A source can be cited for one question and ignored for another. The same product can also be mentioned through a third party even when its own site has strong pages. The measurement method guide helps you keep the source page, question and observed result connected.

How can you measure the gap without fooling yourself?

Keep the question set stable enough to compare like with like, then record the product mention and the cited page as separate fields. A useful report should show the exact answer, whether the product was named, whether a page was cited, which page was cited, and what kind of page it was. It should also preserve the scan date and the question wording. Without those details, a change in the reported result can reflect a different sample rather than a change in visibility.

Use this checklist before you draw a conclusion:

• Record the product mention separately from the linked citation.
• Save the exact buyer question alongside the answer.
• Note whether the cited page belongs to your site or to another domain.
• Label the page type instead of treating every source as equivalent.
• Compare the same questions across scans where possible.
• Read the cited page before deciding that a mention is a success.

The methodology page gives more detail on how to keep those observations consistent. The limit is simple: a scan tells you what appeared for the questions you asked, not what every buyer will see.

What should you change after you find the gap?

If the product is mentioned and another site is cited, do not begin by trying to erase the third party from the answer. Study the page that earned the citation. Check which buyer question it serves, what product facts it states, and whether it compares options in language an engine can reuse. Then make your own page a clearer source for that same decision.

If the product is rarely mentioned at all, citation work alone may miss the problem. A technically accessible page cannot supply recognition if it does not plainly say what the product does, who it serves and how it differs in the situations buyers ask about. In that case, inspect the surrounding pages that establish the product and the facts they repeat.

If your own site is cited but the product is not named, tighten the relationship between the page and the product. Put the product name beside the answer it supports, keep important facts in ordinary text, and make the page useful for the buyer question rather than for a vague category. The aim is to give the engine a clear reason to connect the answer with the product.

That is the moment before the buyer call to notice, not the moment to panic. You have a source trail to inspect and a page decision to make. The free AI visibility checker can help you inspect the difference on your own questions, then you can read the returned answer and source page.

A mention names your product; a citation names the page the engine trusted enough to link.