Key takeaways
- Brand mention rate uses usable answers as its denominator; AI share of voice uses combined appearances of the included tracked brands.
- A competitor can be removed from the comparison and raise your share of voice without changing your brand’s appearances.
- Adding prompts can change both percentages even when your original answer appearances stay the same.
- Brand mentions and website citations are different observations and should not be substituted for one another.
Calculate AI share of voice by dividing your brand’s answer appearances by the combined answer appearances of all brands in your tracked competitor set. Calculate brand mention rate by dividing your brand’s answer appearances by the number of usable answers. The two percentages answer different questions because they have different denominators.
Use the same prompts, platforms, markets, languages and competitor set when comparing periods. Otherwise, a change in the percentage may reflect a change in what you measured rather than a change in brand visibility.
Define the two metrics
For the answer-based method in this article, count a brand once per usable answer, even if its name appears more than once. An answer can count for multiple brands. Exclude missing or unusable results rather than treating them as answers that mention no one.
- Brand mention rate = answers mentioning your brand ÷ usable answers × 100.
- AI share of voice = answers mentioning your brand ÷ total answer appearances across the included tracked brands × 100.
Mention rate asks, “In what percentage of usable answers did we appear?” Share of voice asks, “What percentage of tracked brand appearances belonged to us?” The latter is a share of the selected measurement set, not a share of all AI conversations or customers.
Work through a small dataset
The following data is illustrative, not a real measurement. Assume five usable answers to five tracked prompts and three tracked brands: your brand, A, and competitors B and C.
| Answer | Tracked brands mentioned |
|---|---|
| 1 | A, B |
| 2 | A, B |
| 3 | A, C |
| 4 | B, C |
| 5 | None |
A appears in three answers, B in three and C in two. That is eight tracked brand appearances across five usable answers.
- A’s mention rate: 3 ÷ 5 × 100 = 60%.
- A’s share of voice: 3 ÷ (3 + 3 + 2) × 100 = 37.5%.
The results differ because an answer can mention more than one brand. Answer 5 adds to the mention-rate denominator, but it adds no tracked brand appearances to the share-of-voice denominator.
See what changes when a competitor is removed
Keep the same five answers, but include only A and B in the competitor set. A still appears in three of five answers, so its mention rate remains 60%. Its share of voice is now 3 ÷ (3 + 3) × 100 = 50%.
The increase from 37.5% to 50% comes entirely from removing C’s two appearances from the denominator. It does not mean A appeared in more answers. Record the included competitors beside every share-of-voice result, and recalculate both periods using the same set before comparing them.
See what changes when the prompt set expands
Return to the A, B and C competitor set. Add five illustrative usable answers: two mention only B, one mentions only C, and two mention none of the tracked brands.
| Measure for A | Original five answers | Expanded ten answers |
|---|---|---|
| A’s answer appearances | 3 | 3 |
| Usable answers | 5 | 10 |
| All tracked brand appearances | 8 | 11 |
| Brand mention rate | 60% | 30% |
| AI share of voice | 37.5% | Approximately 27.3% |
A’s count did not fall, but both rates did because the added answers changed their denominators. This comparison describes two different prompt sets; it does not establish that A became less visible on the original five prompts. Report a consistent prompt cohort separately when reviewing a trend.
Make comparisons that can be interpreted
- Set the scope. Record the dates, platforms, markets, languages, prompts, competitor set and number of usable answers.
- Save the counts. Keep each brand’s answer appearances, the usable-answer total and the combined tracked-brand appearances alongside the percentages.
- Compare like with like. Use the same prompts and competitors across periods, or clearly label a scope change and recalculate a comparable view.
- Inspect segments. Break results down by topic or prompt type before drawing conclusions from an overall percentage. When combining segments, use their underlying counts rather than simply averaging their percentages.
- Review the answers. Read the wording around mentions before treating them as recommendations. A mention count alone does not describe the context.
If there are no usable answers, mention rate is undefined. If there are usable answers but no included brand appears, share of voice is undefined because its denominator is zero. Label these cases as unavailable rather than reporting 0%.
Keep brand mentions separate from website citations
A brand name in an answer and a link to a website are different observations. Do not substitute a citation count for a brand-mention count in either formula. For example, Bing’s AI Performance documentation describes total citations as displayed sources in AI-generated answers; it does not describe that count as brand mentions or as placement within an answer. Source
Likewise, check the scope and precision of any external comparison. Bing describes its Topics Portfolio measures as approximate ecosystem signals and says the beta figures should be treated as directional rather than a precise ledger. Those figures should not be assumed to use the same denominator as the answer-based share of voice above. Source
Prepare your AI visibility comparison
- Document the dates, platforms, markets, languages and prompt cohort.
- List the included competitors and apply the same set to each period.
- Save usable-answer and brand-appearance counts with each percentage.
- Recalculate a consistent cohort when prompts or competitors change.
- Inspect answer wording and report citation metrics separately.
FAQs
Is AI share of voice the same as brand mention rate?
No. In the answer-based method shown here, mention rate divides a brand’s appearances by usable answers; share of voice divides them by all included tracked-brand appearances. Neither should be confused with a website citation count, which measures a different observation. Source
Can share of voice rise without more brand mentions?
Yes. In the illustrative dataset, removing a competitor reduces the combined tracked-brand denominator while leaving the brand’s own count unchanged. When using external competitive context, check its scope and precision too: Bing describes its Topics Portfolio ecosystem figures as approximate and directional. Source
Can I use Bing citation counts to calculate brand mention rate?
Not directly. Bing describes total citations as sources displayed in AI-generated answers. The mention rate in this article instead counts usable answers that name a brand, so it requires brand-mention observations and a usable-answer denominator. Source
What should I do when the prompt set changes?
Label the change and compare a consistent prompt cohort separately. Keep the underlying counts so you can see whether a percentage changed because of the measured answers or the denominator. Also check the scope of external query data: Bing says its displayed grounding queries represent a sample of overall citation activity. Source
Discussion
Join the discussion