LLMBrandScan

AI share of voice: the formula and a worked example

By Vladimir Boldyrev · LLMBrandScan · Published July 21, 2026

The formula

AI SoV = (your brand's mentions ÷ total brand mentions in the sample) × 100

It's the same arithmetic as classic advertising Share of Voice, applied to a different kind of “voice” — how often each brand gets named across a sampled set of AI answers, instead of how much ad spend or media coverage each brand commands.

A worked example

Say you run 10 category questions (“best expense-tracking app for freelancers,” etc.) across 5 engines, 3 samples each — 150 total AI answers. Across those 150 answers, brands get mentioned a combined 210 times (some answers name more than one brand; some name none). Your brand is named 42 times.

BrandMentionsAI SoV
You4220%
Competitor A6330%
Competitor B4622%
Everyone else5928%

42 ÷ 210 = 20% AI Share of Voice — behind Competitor A's 30%, ahead of Competitor B's 22%. That single number is a snapshot; the useful version reports it as an observed range (e.g. “18-23% across this sample”) with a Strong / Emerging / Noisy signal label, since a different sample of 150 answers would land somewhere nearby but not identical.

Where the concept comes from

AI Share of Voice is the percentage of brand mentions, across a set of sampled AI answers in a category, that belong to a given brand — calculated as that brand's mention count divided by the total mention count of every brand named across the same sample, summing to 100% across the full competitive set. It is a direct adaptation of the classic marketing metric Share of Voice, and echoes Les Binet's Share of Search (a brand's search-query volume divided by total category search volume), which the IPA's EffWorks research showed could predict market-share movement roughly a year ahead for some categories. AI Share of Voice inherits that positioning: a fast, sampled, behavioral proxy for standing in a category — not a survey of what every buyer sees, and not a claim about actual market share, only a leading indicator drawn from AI answers.

Why this metric specifically is worth tracking

Unlike raw mention count, Share of Voice is inherently comparative — it tells you your standing relative to competitors in the same category, not just whether you showed up at all. Paired with first-mention rate (which brand a model names first), it's the closest AI-native equivalent to the “excess share of voice predicts future market-share growth” relationship classic marketers already rely on. See a full worked report in our sample scan.

See where you actually stand — not a guess.

LLMBrandScan runs your category's questions live across up to 5 AI engines and reports the observed range across samples, with a Strong / Emerging / Noisy signal label — not a single lucky answer, and no promise about where you'll rank.

$19 = 90 live answers across 3 engines · no subscription · ranges, not fake precision

More AI-visibility guides.