Why can two platforms report different share-of-voice numbers from the same data?

Because every word in the definition matters — what counts as a mention, which competitors qualify, and which prompts and providers are included all change the result. Publish the calculation beside the number itself; without it, two platforms can show different percentages from the same underlying answers and both appear correct.

Is being mentioned the same as being recommended?

No, and merging them hides real signal. Presence tells you whether a brand appeared at all, prominence captures where and how strongly it appeared, and recommendation distinguishes a brand being listed from a brand being explicitly selected as the best fit — these are three different things worth tracking separately.

What happens to share of voice if I change my competitor list?

Changing the competitor set changes the denominator, so the number itself shifts even if nothing about your actual visibility changed. Record when competitors are added or removed, and avoid comparing periods with materially different scopes without noting the change.

My share of voice moved — what should I check first?

Don’t stop at whether it went up or down. Ask which specific prompts changed, which provider produced the movement, who gained when you lost ground, and which citations appeared in the winning answer — that’s what turns a percentage into something actionable.

Share of voice in AI answers: what should you measure?

Mention share is useful only when the prompt set, competitors, provider coverage and calculation are visible.

KEY TAKEAWAYS

The short version

✓Define the denominator before publishing share of voice.

✓Separate presence, prominence and recommendation.

✓Use a stable competitor set for trend comparison.

✓Always show provider and prompt coverage.

01 / 04

Start with a transparent definition

A basic AI share-of-voice metric divides a brand’s qualifying mentions by all qualifying mentions across the monitored brand and competitor set. That definition sounds simple, but every word matters: what counts as a mention, which competitors qualify, and which prompts and providers are included?

Publish the calculation beside the result. Without it, two platforms can show different percentages from the same answers and both appear correct.

02 / 04

Measure more than presence

Presence tells you whether a brand appeared. Prominence captures where and how strongly it appeared. Recommendation distinguishes a brand being listed from a brand being selected as the best fit. These signals should not be silently merged.

●Mention rate: share of monitored answers containing the brand.

●Mention share: brand mentions divided by tracked-brand mentions.

●First-position share: frequency of being the first named option.

●Recommendation share: frequency of receiving an explicit endorsement.

03 / 04

Keep the comparison set stable

Changing the competitor list changes the denominator. Record when competitors are added or removed and avoid comparing periods with materially different scopes without a note.

Segment by market or buyer intent when the same competitors do not operate across every category. A single blended number can hide where a brand is genuinely winning.

04 / 04

Use share of voice as a diagnostic

The useful question is not only whether share increased. Ask which prompts changed, which provider produced the movement, who gained when the brand lost, and which citations appeared in the winning answer.

TURN THE FRAMEWORK INTO EVIDENCE

See what the models say about your brand.

Monitor buyer questions, raw answers, citations and search performance in one traceable workspace.

Book a demo ↗