The job: one basis, every surface
Buyers ask the same questions in different places: Google's AI results, Perplexity, search-grounded assistants. A cross-surface tracker runs one fixed question set across all of them on the same schedule, so the comparisons between surfaces are real rather than remembered.
That shared basis is the whole value. Two surfaces checked with different questions on different weeks produce anecdotes that cannot be compared, no matter how good the dashboard looks.
Why the blended score is a trap
Averaging visibility across surfaces manufactures a number with no owner and no action. Being strong on Perplexity and absent from Google AI results averages to mediocre, which describes neither fact and hides the one that needs work.
Surfaces also move for different reasons: retrieval changes shift one, a model update shifts another. A blend cannot be diagnosed; a per-surface split diagnoses itself.
What to share and what to keep separate
| Topic | Element | Shared or per surface | Why |
|---|---|---|---|
| Question set | Shared | One basis makes cross-surface comparison legitimate and keeps history comparable. | |
| Mention rate | Per surface | Each engine answers from different data; the rates are different facts. | |
| Citations | Per surface | Each surface draws on its own source mix; the levers differ accordingly. | |
| Trend history | Per surface | Movements have surface-specific causes and need surface-specific responses. | |
| Cadence and sampling rules | Shared | Same repetition and schedule everywhere keeps the method honest. |
What the weekly read should contain
- Mention rate per surface with its sample size and confidence range, no averages across surfaces.
- The prompts where any surface changed beyond normal variation, with the samples to prove it.
- Citation shifts: domains that entered or left the answers on each surface.
- Competitor deltas on the shared basis, per surface.
- Any change to the question set, marked, so next week's read stays comparable.
Where AeoWatch fits
AeoWatch is built as exactly this instrument: one tracked question set, supported AI surfaces measured separately including Google AI results, per-surface mention rates with confidence intervals, and citations recorded per answer. The AI search tracking tools guide covers the wider category if you are still comparing options.