Visibility score vs share of voice: which number should you report?
Visibility score and share of voice come from the same AI answers but answer different questions. How each is calculated, when they disagree, and which one belongs in front of leadership.
Visibility score and share of voice are calculated from exactly the same AI answers. They often move together. Then one month they don't, and someone in a leadership meeting asks which one is true.
Both are. They answer different questions, and the difference comes down to what sits underneath the fraction.
Two numbers, two denominators
Visibility score is the share of collected answers that name your brand. Answers naming you, divided by answers collected. The denominator is fixed by your prompt set and the number of checks.
Share of voice is your mentions as a share of all mentions of the brands you track. Your mentions, divided by your mentions plus every tracked competitor's mentions. The denominator depends on how often your competitors are named.
That one difference explains every case where they disagree.
The score answers: when a buyer asks one of these questions, how likely are they to see our name? Share of voice answers: of the attention AI answers give to this group of brands, how much goes to us?
A worked example where they split
All numbers here are invented for illustration. You track a set that produces 100 answers a month, and you watch two competitors, A and B.
| Month | You named | A named | B named | Your score | Your share of voice |
|---|---|---|---|---|---|
| 1 | 30 | 50 | 40 | 30% | 25% |
| 2 | 30 | 20 | 40 | 30% | 33% |
| 3 | 30 | 20 | 40 | 30% | 23% (after adding competitor C, named 40 times) |
In month 2, your share of voice jumped eight points and nothing happened to you. Competitor A lost ground, maybe because an engine stopped citing one of its pages. In month 3, someone added competitor C to the tracking set, and your share of voice fell below where it started. Again, no answer changed for you.
Your visibility score sat at 30% throughout, because buyers saw your name exactly as often as before.
The opposite split happens too. If engines start giving longer answers that name more brands, your score can rise while your share of voice falls, because everyone is getting mentioned more.
Counting rules that change share of voice
Share of voice needs a few more decisions than the score, and each one changes the result.
Mentions per answer. If an answer names a competitor three times and you once, is that 3 to 1 or 1 to 1? Counting once per answer is simpler and harder to distort. Counting every mention rewards brands an answer dwells on.
Position. Some teams weight the first brand named more heavily than the fifth. That can be reasonable, but it is a model of buyer attention, not a measurement, so write the weights down and keep them fixed.
Who is in the set. This is the big one. Share of voice ignores every brand you do not track. If engines keep recommending three companies that are not on your list, those mentions do not exist as far as share of voice is concerned. You can hold a comfortable share of voice against your chosen rivals while buyers are being pointed at someone else entirely.
That last problem has a simple guard. Once a month, read a sample of answers to your weakest prompts and list every brand named. If an untracked brand keeps appearing, discuss adding it, and log the change the same way you would log a prompt change.
When share of voice is the better number
Share of voice earns its place in a few situations.
In a crowded category where you are already named in most answers, the score stops telling you much. The interesting question becomes whether you are one of five names or one of two. Share of voice shows that.
At the prompt level, share of voice shows who you are losing to. A prompt where you are named 1 time in 4 checks and competitor A is named 4 times in 4 tells you exactly whose pages and citations to study.
And in competitive reviews, share of voice is the natural format, since the whole point is the comparison.
Why the score is our headline
We lead with the visibility score because it does not depend on the competitor list and it is closest to what matters commercially: a buyer reading the answer and seeing your name. Nobody can move it by editing a spreadsheet of competitors. It also answers the question leadership usually asks, which is "are we showing up", not "how do we compare with three named rivals".
Share of voice sits directly underneath, against a competitor set agreed at the start of the engagement. If the two disagree, the report says why in one sentence. Usually it is one of three things: a competitor gained or lost, the competitor set changed, or answers got longer or shorter.
Read the answers, not just the counts
Both numbers treat a mention as a mention. An answer that says "avoid [Brand], the support is poor" counts the same as a recommendation. So alongside both numbers, look at the sentiment of the answers that name you, and read the ones that name you in a negative context. A share of voice built on warnings is not a win.
How we measure both in Promptwatch
Both numbers in our reports come out of Promptwatch. Prompt tracking gives the visibility score per engine. Competitive benchmarking and share of voice are calculated against the competitor set we configure for each client, so changing that set is a deliberate act rather than a side effect. Sentiment analysis is attached to the responses, and because the responses themselves are stored, we can open any prompt and see which brands were named, in what context, and which pages were cited for it.
When the score and share of voice disagree, Agent Chat is a quick way to ask why over the live data, for example which competitor gained the most mentions this month and on which prompts. We still check the answer against the stored responses before it goes in a report, but it shortens the search a lot.