The math behind the score.
The AI Visibility Score is a composite 0 to 100 metric. Four signals, weighted, normalized across every engine you track. Here is exactly what goes in and what does not.
What goes into the score.
Each signal answers a different question a CMO would ask. The composite blends them so a single number captures presence, prominence, and authority.
Mention rate
How often your brand is named in AI responses to the queries you track. Computed per engine, then averaged. The baseline question: are you even in the answer.
Share of voice
Of the answers that mention any brand in your category, how often is yours the lead, also-mentioned, or absent. Lead beats also-mentioned. Also-mentioned beats absent.
Average position
When you are cited, where do you appear in the answer. First-listed brands carry more weight in buyer recall than mid-list ones. Position scores normalize across answer length.
Citation share
Of the cited answers in your category, what fraction include a citation to your domain. Citation share separates being talked about from being authoritative.
A score, line by line.
Composite signals roll up to a single number. Here is what that looks like for one engine, one week, mid-market SaaS observability category.
Inputs · ChatGPT · Week 12
| Signal | Value | Weight |
|---|---|---|
| Mention rate | 54% | 35% |
| Share of voice | 41% | 30% |
| Avg position | 3 of 5 | 20% |
| Citation share | 22% | 15% |
Roll-up · weighted composite
| Step | Value |
|---|---|
| 0.35 × 54 | 18.9 |
| 0.30 × 41 | 12.3 |
| 0.20 × (position-normalized 60) | 12.0 |
| 0.15 × 22 | 3.3 |
One score, nine surfaces.
Each engine produces its own composite. The unified AI Visibility Score blends them, weighted by the share of category traffic each engine carries for your geography.
Honest about the edges.
A composite score is useful because it simplifies. It is also useful to be clear about what it leaves out, so the number you report is the one you can defend.
Sentiment of the mention. The score measures presence and prominence, not whether the model framed you positively. Sentiment is reported separately, not folded in.
Click-through or downstream conversion. The score measures what the answer says about you, not what the buyer does next. Pipeline attribution lives in your CRM, not in Zumi.
Paid placement or sponsored mentions. Where an engine surfaces paid units alongside organic answers, only the organic answer is scored.
Model updates change behavior. When an engine ships a new model version, the score can shift even with no change in your content. Zumi flags these events inline on the trend chart.
Your competitors are already paying attention.
Zumi shows you exactly where your brand stands in AI-generated answers, across every engine, against every competitor. Book a 30-minute working session and we'll show you live data for your brand. No deck. No pitch.