Quick answer
Mention rate shows whether you are in the conversation, position shows how you are treated when named, and share of voice shows how you compare with competitors. Read together, with sample sizes, they point to different work.
It is tempting to collapse AI visibility into one score. It hides more than it shows. Three simple measures, read together, tell you far more.
Mention rate: are we in the conversation?
The share of eligible answers that name your brand. It answers the basic question of presence. A low rate means you are absent from the answers your customers are reading.
Position: how are we treated when named?
Where you appear when an answer lists several options, and how you are described. Being named last with a weak description is different from being named first with a clear one.
Share of voice: how do we compare?
Your mentions divided by all tracked brand mentions in the same answers. It answers whether you are ahead or behind the competitors you chose, on the same prompts.
Reading them together
Always show the sample
Whichever measure you use, show how many observations sit behind it. Small samples swing widely, and honest reporting says so.
How each number is calculated
Precision about definitions is what makes the three numbers useful. Mention rate is the count of eligible answers that name your brand divided by the count of eligible answers. An answer is eligible when the run completed and the prompt is one where your brand could reasonably appear. Position is measured only on answers where you are named, and it describes where you appear in a list and how you are described. Share of voice is your mentions divided by the total mentions of all brands you track across the same answers.
The denominator is where reports go wrong. If failed runs are counted as misses, the rate falls for reasons that have nothing to do with your brand. If the competitor list changes between periods, share of voice moves without any change in how AI treats you. Fix the definitions once and keep them fixed.
Reading the three together
| Mention rate | Position when named | Share of voice | Likely reading |
|---|---|---|---|
| Low | Strong | Low | You are excellent when found and rarely found. Work on presence: coverage of the topics, sources that mention you, retrievable pages. |
| High | Weak | Mid | You are known but undersold. Work on how you are described: clearer statements of what you do and who it is for. |
| High | Strong | Low | You do well, but a rival or a group of rivals appears far more often. Look at which prompts and sources they own. |
| Low | Weak | Low | Start with the basics: is the site crawlable, are the facts consistent, does anyone independent describe you. |
Choosing the prompts behind the numbers
All three numbers inherit the quality of the prompt set. A set made only of branded questions will show a flattering mention rate that says little about discovery. A set made only of broad category questions will show a low rate that may be unfair to a niche brand. The most informative sets mix discovery questions, comparison questions and questions about specific needs, and they report each group separately before combining.
- Discovery prompts test whether you are named when a buyer does not know you yet.
- Comparison prompts test how you are framed against named alternatives.
- Need-based prompts test whether you appear for a specific use case, price band or location.
- Reputation prompts test what is said about you when a buyer asks about your brand directly.
Common mistakes
- Reporting a single blended number across platforms, which hides which assistant is failing.
- Comparing a period with 20 prompts to a period with 60 and calling the difference progress.
- Treating position as a fixed rank. AI answers have no stable ranking, so position describes a pattern over many answers.
- Ignoring the sample size, then reacting to movement that is within the noise.
What to put in front of leadership
Which of the three should I track first?
Start with mention rate, because presence comes before treatment. Add position and share of voice once you have a stable prompt set and enough repeat runs to trust the trend.
Can share of voice exceed 100 percent?
No. It is your mentions as a share of all tracked brand mentions in the same answers, so it sits between zero and one hundred percent. It changes when the competitor list changes, which is why the list should stay fixed between periods.
Sources
- 1. BrandWater AI: Methodology (Current)
Cite this page
BrandWater AI Research. (2026, 8 September 2026). Mention rate, position and share of voice: three numbers, three questions. https://brandwaterai.in/blog/three-numbers-three-questions
How we work
Figures are dated and linked to their sources. Where none exist we say so. Read our methodology and AI transparency pages.