The short answer
Really Good GEO records an AI mention when the visible answer names the brand. It records an AI citation when the interface visibly attributes information to a source URL or domain. A response can contain a mention without a visible citation, a citation without a brand mention, both, or neither.
This is a measurement taxonomy from Really Good GEO. Product interfaces differ, so record the observable result instead of assuming every product exposes sources in the same way.
Keep four result states separate
| Visible answer | Record | Do not infer |
|---|---|---|
| Brand named and brand page cited | Mention: yes. Citation: yes. | That the result will persist in another run. |
| Brand named with no visible source | Mention: yes. Citation: no visible citation. | Which source or retrieval process produced the mention. |
| Another page cited but brand not named | Mention: no. Citation: record the visible source. | That the cited page created brand awareness. |
| No brand and no visible source | Mention: no. Citation: none visible. | That no retrieval occurred. |
A citation is not the same as retrieval
Retrieval describes information being brought into the system's working context. A visible citation is an interface-level attribution the observer can record. If the product does not show a source, the test does not establish whether retrieval occurred or which material informed the answer.
Record enough context to repeat the observation
- Exact question and question type
- Named AI product
- Date and run number
- Whether visible web search occurred
- Whether the brand was mentioned
- Every visible cited domain and URL
- Competitors surfaced
- Page version or controlled change
The free AI Citation Test Log keeps these fields together.
Report mentions and citations as separate measures
For a repeated question set, summarize mention observations and visible-citation observations separately. Keep brand-free, comparison, and known-brand questions separate as well. Publish the product, dates, run count, question set, exclusions, and limitations with any result.
Start with the page being tested.
Audit the source page for clarity, evidence, entity context, and quotability before changing it.
Audit a page free