An AI answer can name your brand and steer the buyer away in the same sentence. How to score what a mention is actually worth.
Most AI visibility dashboards open with one number: how often your brand is named. It is easy to compute and easy to chart. It is also the number most likely to rise while you lose buyers.
The problem with counting mentions
A buyer asks an assistant for the best tool in your category for a mid-sized team. The answer names four vendors. You are one of them — listed third, with a note that you are “better suited to enterprises with dedicated admins”. Counted as a mention, that answer is a win. Read the way a buyer reads it, it is a polite exclusion.
A mention count treats three very different answers as the same event:
- You are named as the one to choose.
- You are listed, with no view expressed either way.
- You are named, and the buyer is advised against you.
Grade the endorsement, not the appearance
The fix is to grade how each answer treats you and let that grade decide what the mention is worth. Vidrys reads every answer against a five-level scale:
| Level | What the answer does | Worth |
|---|---|---|
| Strong | Names you as the one to choose | 1.0 |
| Recommended | Names you among the options worth considering | 0.8 |
| Neutral | Lists you with no view either way | 0.5 |
| Reference only | Cites you as a source, not a choice | 0.2 |
| Discouraged | Names you and advises against you | 0.0 |
The grade multiplies the visibility score rather than sitting beside it. A dismissive mention contributes nothing, however prominently it appears.
Position still matters
Among answers that do recommend you, order matters: buyers read shortlists from the top, and the first names in a list are the ones people remember — the serial-position effect. Position — first pick, considered alternative or footnote — is worth tracking as its own number, next to share of voice and citation share. The GEO study that named generative engine optimization (Aggarwal et al., 2023) measures a source’s visibility in an answer the same way: by how much of the answer it accounts for, weighted by how early it appears.
What changes when you score this way
- The score stops rising on answers that hurt you.
- Losses become specific: you can see which answers named you and still sent the buyer elsewhere.
- Fixes target the reason you were discounted, not just your absence.
Where to start
Pull the answers for your ten most commercially important prompts and read them as a buyer would. Mark each one strong, recommended, neutral, reference-only or discouraged. The gap between your mention rate and your endorsement-weighted score is the size of the problem. Vidrys scores every answer this way on every plan, from Solo at ₹999 a month. For the full method, read the GEO & AEO guide.
Written by the Vidrys team. Published 13 Sept 2026.