The Visibility Index
The index is deliberately simple so it can be explained in one sentence: it rewards being recommended, and being recommended first.
Per answer (a run)
Section titled “Per answer (a run)”Each run is one prompt asked to one provider in one mode and one language. From the answer we extract the recommended brands in order. Your brand’s contribution is:
position_weight = 1 / position of your first mention (0 if you are absent)First place scores 1, second 0.5, third 0.33 — absence scores 0.
Per slice
Section titled “Per slice”A slice is a combination of provider, mode and language; the blended slice is all runs together. For every slice and day:
| Metric | Definition |
|---|---|
| Visibility Index | 100 × mean(position_weight) over the slice’s runs |
| Mention rate | 100 × runs where you are mentioned / all runs |
| Average position | mean position of your first mention, over runs where you appear |
| Share of voice | 100 × your weighted mentions / all tracked brands' weighted mentions — see Share of voice |
So an index of 50 could mean “first in half the answers, absent in the other half” or “second everywhere”. Mention rate and average position tell the two apart.
Why an average of everything
Section titled “Why an average of everything”Because it makes days comparable. Adding a language or a question changes the denominator, so the blended trend is most meaningful when the tracked set is stable; the per-slice views isolate a single assistant, mode or language when you need to know where a move came from — and the alerts do that for you.
What counts as a mention
Section titled “What counts as a mention”Extraction is done by a model reading the answer and listing recommended brands with their position and framing (positive, neutral, negative). Your brand’s aliases (set in the brand profile) are matched; brands recommended that you do not track are kept as unmatched and surfaced as “brands appearing in your market”.