Share of voice vs share of answers
Guide · Buyer Research & Comparisons · 4 min read · last verified 2026-07-27
Share of voice is a family of metrics estimating how much of a market's visible conversation belongs to a brand — classically its slice of media coverage, advertising presence, or social mentions over a period. Share of answers measures something narrower and newer: across a defined set of buyer questions, the fraction of actually delivered answers — from AI assistants and answer-style search surfaces — in which the brand appears. The names sound like siblings. The metrics are not: they count different events, over different denominators, in service of different decisions.
Two denominators, two different claims
Every share metric is a fraction, and the denominator decides what the fraction can claim. Classic share of voice divides your mentions or impressions by the market's total conversation volume in chosen channels over a time window. That denominator is channel-shaped and elastic: it swells when the category gets loud and shrinks when it goes quiet, whether or not a single buyer changed behaviour.
Share of answers divides differently. Fix a set of questions buyers raise while deciding what to buy — building that set well is its own discipline, treated in what is a buyer question map — collect the answers actually returned for each, and count the ones that include you. The denominator is question-shaped and deliberately held fixed, which is what makes movement interpretable: if the set is stable and your presence rises, something real changed upstream of the answers.
Because the denominators differ, the numbers do not convert. A brand can dominate the visible conversation yet be missing from composed answers; a quiet brand can be an assistant's steady recommendation. Averaging the two, or quoting one as evidence about the other, produces a figure with no referent.
What classic share of voice honestly supports
Share of voice earned its seat and keeps it. It is the natural metric of media and brand work: campaign lift, coverage weight, share of category noise. Its history is an asset in itself — teams have tracked it for a long time, so internal trend lines are long, definitions are settled, and quarter-over-quarter comparison is meaningful.
It also covers something per-question presence never will: reach into people who are not buying anything yet. Much of what share of voice measures lands on the uncommitted, and building memory in that audience is a legitimate, slow-payoff goal that no answer-presence number can proxy. When the decision at hand is media allocation, agency accountability, or brand tracking, share of voice is the fit-for-purpose gauge, and swapping it out would be a mistake.
What share of answers measures — and its sharp edge
The sharp edge is resolution. The unit is the single question, so the metric does not merely report a level; it localises the problem. It can say which questions you are absent from, which answers your rivals occupy, and — in well-built implementations — what evidence the winning answers cite. That per-question grain is what turns the number into a work queue rather than a mood.
Two caveats keep it honest. Composed answers are volatile: they can shift between scans without any action from you, which is why measurement cadence and locked baselines matter — see how often should you re-scan AI visibility. And the level alone is not self-interpreting. What counts as healthy, and how to read movement over time, is a separate question with its own treatment in what a healthy AI share of voice looks like; this piece stays at the definitional layer.
The conversion mistake
The commonest error in practice is inferring one metric from the other. A rising share of voice gets read as evidence that answer presence must be improving too — but the pipelines are different. Mentions accumulate through coverage and campaigns; composed answers weigh documentation, comparison content, and community evidence by their own logic, which a burst of coverage may never touch. The audiences differ as well: an impression lands broadly on people who asked for nothing, while an answer lands on one buyer at the exact moment of a question.
The error runs the other way too. A team winning answers can conclude that brand investment is waste — forgetting that answer presence says nothing about the memory being built in people who have not started asking questions yet. Each metric is silent exactly where the other speaks.
Keeping both without confusing them
| Share of voice | Share of answers | |
|---|---|---|
| Counts | Mentions, impressions | Appearances in delivered answers |
| Denominator | Channel conversation volume, elastic | Fixed buyer-question set |
| Resolution | Aggregate, by channel | Per question |
| Moves when | Coverage and campaigns move | Sources and answer composition move |
| Decision supported | Media planning, brand tracking | Content priorities, gap-closing |
| Silent about | Composed answers | Pre-buying awareness |
Report them side by side, each against its own decision, and never blend them into a single visibility score. Magrios operationalises the second column — holding the question set fixed and re-scanning the answers so change is attributable — while leaving the first column to the tools built for it. For the practice-level version of this contrast, watching mentions of yourself versus measuring the whole buying conversation, see brand monitoring vs Market Growth Intelligence.