Cited vs. Mentioned: The Two Types of AI Visibility

Not every time an AI model says your brand name is the same kind of win. Sometimes it recommends your product with a real, verifiable source behind it. Sometimes it just recalls your name from what it already knows, with nothing to back it up. Both look similar in a transcript — but only one of them is a citation.

Understanding the difference is the first step to reading your AI Visibility results correctly, instead of celebrating (or panicking over) the wrong number.

The Two Signals AI Models Send

Cited — the AI model’s answer includes a real, checkable source that points to your domain. The model retrieved something — a page, a product listing, a piece of content — and used it to build part of the answer. This is the equivalent of a backlink: traceable, verifiable, and it can drive a click.

Mentioned — your brand name shows up in the answer text, but there is no source attached to it. The model recognized your name from what it already knows and used it, without retrieving or verifying anything current. It is a real signal — the model has heard of you — but it proves recognition, not recommendation.

	Cited	Mentioned
What happened	Model retrieved and used a real source	Model recalled your name from training data
Verifiable	Yes — a real URL or source is attached	No — just text
Can drive traffic	Yes	No
What it proves	Your content was used to answer a real query	Your brand is known, but not necessarily current
Strength as a signal	Strong	Weak, but not nothing

Want to know whether AI engines actually cite your domain or merely recall your brand? Check your brand across four AI engines for free →

Why the Difference Matters

Treating every brand mention as a “citation” inflates your numbers and hides where the real work is. A store that gets mentioned often but cited rarely has a recognition problem solved and a content problem unsolved: people know the name, but the AI has nothing current and specific enough to actually recommend it with a source.

The reverse is rarer but healthier: fewer mentions, more citations, means the content that does exist is doing real work — getting retrieved and used, not just recalled.

What Causes Each One

A mention usually comes from training data — the model has encountered your brand before (in reviews, press, other sites) and associates it with a category or topic, but has nothing fresh or specific enough to cite for the current query.

A citation requires the model to retrieve something at answer time — through live search or a grounding step — and find it specific and structured enough to point to. That means: a page exists that directly answers the question being asked, the claims on it are specific enough to be worth citing (not vague marketing language), and the page is technically reachable by AI crawlers in the first place.

How to Move From Mentioned to Cited

Make claims that are worth citing. “High quality” is not citable. “IP67 water resistance rated to 1 meter for 30 minutes” is. Models cite specifics because specifics are verifiable.

Answer the exact question, not the general topic. A page about your brand in general is weaker than a page that answers “what is the best X for Y” directly and completely.

Keep structured signals current. Schema.org markup and an llms.txt file give AI crawlers a clean, unambiguous way to find and parse your most citable content — without it, even good content can go unretrieved.

Publish content the model can actually reach. If AI crawlers are blocked, throttled, or the content lives behind heavy client-side rendering, there is nothing to retrieve — no matter how good the underlying page is.

None of this happens overnight. A mention can exist from old training data with zero current content behind it; a citation requires something live, specific, and reachable right now.

How to Track Both

Most tools that check “AI visibility” only look for your brand name in a response and call it a win. That conflates the two signals we just described and gives you a number that looks better than your real citation rate.

Krawfly’s AI Visibility Score and Citation Check track cited and mentioned as two separate signals for every product — so you know exactly which one you are getting, and where the gap is between “known” and “recommended.”

If another brand has the linked source while yours is only recalled by name, use the workflow in Why ChatGPT Cites Your Competitors. For a complete baseline covering readiness, queries, evidence, and persistence, follow the AI Visibility Audit for E-commerce.

Check Your Brand

Krawfly checks what ChatGPT, Claude, Gemini, and Perplexity know about your brand. The result keeps available signals separate, so a simple mention is not mistaken for a citation.

Run the free AI Brand Check →

Frequently Asked Questions

Is a mention still valuable? Yes, but as a recognition signal, not a traffic or recommendation signal. It tells you the model has encountered your brand before — useful context, not a result to optimize toward on its own.

Can a mention become a citation over time? Yes. As you publish more specific, structured, and citable content, and as models retrain or re-index, brand recognition can convert into real citations — but it requires content that gives the model something concrete to point to, not just more mentions.

Why do some AI engines cite more than others? It depends on whether the engine performs live retrieval (search-grounded) for a given query or answers purely from training data. Search-grounded engines cite more often because they are actively retrieving sources at answer time.

Should I ignore mentions entirely and only track citations? No — track both, but weight them differently. A rising mention rate with a flat citation rate tells you exactly what to fix next: your content needs to become citable, not just findable.