AI visibility becomes useful when you can explain what was measured. Two of the clearest signals are mentions and citations. They are related, but they answer different questions.
- A mention asks: did the AI answer name the business?
- A citation asks: did the AI platform provide the business website as structured source evidence?
Keeping these signals separate makes reporting clearer and prevents a visible brand name from being mistaken for source-backed visibility.
What counts as an AI mention?
A mention is a recognized business name inside the saved AI answer. Measurement should use the configured business name and its approved variations—not loose words that happen to look similar.
Imagine a business called Northstar Health with an approved variation of Northstar Clinic:
- “Northstar Health offers preventive-care programs” is a mention.
- “Northstar Clinic is one option to consider” is also a mention.
northstarhealth.exampleappearing only as a URL is not a name mention.
That last distinction matters. A domain inside a link should not automatically become evidence that the answer actually discussed the brand.
What counts as an AI citation?
A citation comes from the structured sources returned by the AI platform. The source domain must match the tracked business website or one of its subdomains.
For example, a result can count as a citation when the provider returns:
northstarhealth.exampleresearch.northstarhealth.example
A domain typed into ordinary answer text is not enough by itself. The provider must return it as source evidence.
One answer can produce four outcomes
| Mention | Citation | What it means |
|---|---|---|
| Yes | Yes | The answer names the business and cites its website. |
| Yes | No | The business is discussed, but its website is not supplied as a source. |
| No | Yes | The website supports the answer, even though the business name is not written in the answer. |
| No | No | The answer neither names the business nor cites its site. |
None of these states should be inferred from a single score. Saving them independently shows exactly what changed between platforms and over time.
Why evidence matters
A percentage without the underlying answer and sources is difficult to defend. If a client asks why citation visibility changed, the useful response is not simply “the score went down.” You should be able to inspect:
- The exact prompt that was measured.
- The AI platform and configuration used.
- The saved provider answer.
- The exact brand-name span that produced a mention.
- The structured source that produced a citation.
This evidence also protects historical reporting. If a business later changes its name, website, or tracking setup, previous results should remain connected to the identity and configuration used for that run.
Measure both, then compare consistently
Mentions describe whether the brand appears in the answer. Citations describe whether its website appears in the provider’s source evidence. Neither replaces the other.
The practical approach is to measure both against repeatable prompts, retain the evidence behind each result, and compare like-for-like runs across AI platforms and over time.