Most brands tracking AI visibility are counting the wrong thing.
They run a few prompts, see their brand name in a ChatGPT response, and report back to leadership that “we are appearing in AI search.” What they are not reporting is whether that appearance was a citation that signals content authority, a passing mention in a comparison list, or an unverified claim the AI generated from training data that may not even be accurate.
Those three outcomes are not the same. They require different fixes. And without a clear measurement framework, you cannot tell which one you are actually getting.
What a Mention Actually Is
A mention is when your brand name appears anywhere in an AI-generated answer. It is the lowest bar of visibility: AI systems are aware your brand exists and have placed it in context relevant to a query. A mention does not require a link. It does not require accurate information. It does not mean the AI is treating your content as authoritative.
Example: “Popular options in this category include Brand A, Brand B, and Brand C” is three mentions. None of those brands is being cited as a source. The AI is naming them because it has encountered them in training data or retrieved content, not because any one of them produced something the AI is specifically pointing to.
What a Citation Actually Is
A citation is when your domain appears in the source reference panel of an AI response, meaning the AI retrieved your content, used it to generate or support an answer, and pointed to it as the source of a specific claim or recommendation. This is qualitatively different from a mention because it means your content was considered trustworthy enough to function as the basis for the AI’s answer.
On Perplexity, citations are visible as numbered inline references and a source panel. On ChatGPT with web search enabled, they appear as clickable links. On Google AI Overviews, they appear as source cards below the generated summary. On Claude and Gemini, citation visibility varies by mode.
A linked citation is stronger still: not only was your content cited, but the AI provided a clickable link that could drive actual traffic back to your site. This is the only AI visibility outcome that has a direct, measurable traffic impact, which is why tracking linked versus unlinked citations separately matters for any team trying to connect AI visibility to business outcomes.
The Gap Between Them Is Diagnostic
Here is where it gets strategically useful. The relationship between your mention rate and your citation rate tells you exactly what kind of visibility problem you have.
High mentions, low citations: AI systems know your brand exists but do not treat your content as authoritative enough to source. This is typically a content structure problem. Your pages are not organized for passage-level extraction. Your key claims are buried in narrative prose instead of being stated directly in the first sentence. You are missing FAQ-structured sections, named statistics with sources, or proper schema markup. The fix is structural, not promotional.
Low mentions and low citations: Your brand has weak entity recognition across AI platforms. The models do not reliably identify you as a distinct brand in your category. This is an entity problem: inconsistent representation across third-party sources, Wikidata, and public databases. Fixing this requires building off-site presence, not just improving your own pages.
High citations, low traffic: Your content is being cited but the citation is unlinked, meaning your brand name is referenced without a clickable link. You are building AI brand awareness without capturing the traffic that could convert it. Understanding which of your citations are linked versus unlinked is essential before you can diagnose this correctly.
This is exactly why the difference in citation behavior across ChatGPT, Gemini, and Perplexity matters so much: the same brand can have radically different mention and citation rates on different platforms, and treating them as a single number obscures the specific fix each situation requires.
Why Most Teams Are Measuring This Wrong
The most common mistake is collapsing mentions and citations into one undifferentiated “AI visibility” number. A tool or report that tells you “your brand appeared in AI answers 1,200 times this month” is not telling you what that appearance meant. Were they linked citations in Perplexity that drove actual clicks? Unlinked passing mentions buried at the end of a long list? Neutral comparisons where your brand was listed third after two competitors who were described more positively?
A 1,200-mention month could represent a genuine competitive advantage or a complete waste of potential visibility, and you cannot tell which without separating the signal.
Platforms like Authority Radar are built specifically to make this distinction: tracking linked citations, unlinked mentions, citation position, and sentiment separately, across eight or more AI platforms, with 99 percent citation accuracy using a citation-first detection method rather than the domain-rollup approach that many simpler tools use. Domain-rollup detection checks whether your domain appeared anywhere in a response. Citation-first detection tracks the specific citation event, its position, and whether it carried a link, which is the data you actually need to act on.
What to Do With the Distinction
Once you are measuring mentions and citations separately, the path forward becomes clearer.
If you have a citation gap, which is where a competitor consistently gets cited for queries where you are absent or only mentioned, the fix is almost always content-based: building pages that directly answer the specific questions driving those queries, structured so AI systems can extract a passage and cite it confidently. The framework for finding and closing AI citation gaps walks through this process in detail.
If you have a schema gap, where your content exists but is structurally hard for AI systems to parse, schema markup implementation is often the highest-leverage single change you can make.
In both cases, the starting point is the same: accurate measurement that separates what AI is saying about you from whether AI is actually citing you. Start your free trial and get both numbers, by platform, for your brand.
