# Share of answer is the new share of voice

Published: 2026-07-25
Author: Brandflare Team (https://brandflare.io)
Canonical: https://brandflare.io/blog/share-of-answer-is-the-new-share-of-voice

**The short answer:** share of voice asked "of all the conversation in our category, how much mentions us?" Share of answer asks the sharper version: "of all the AI answers to our customers' questions, how many name us?" It's countable, trendable, and — unlike a lot of brand metrics — directly tied to a shortlist a real buyer sees.

## Why the old metric doesn't stretch

Share of voice was built for media you could sample: press mentions, social posts, search impressions. AI answers break the sampling model in two ways. First, they're generated fresh per conversation — there's no archive to count. Second, they're winner-take-most: an answer names three brands and omits thirty, with no long tail of page-two visibility to partially credit. You can't survey what customers were told; you have to ask the same questions yourself and count who shows up.

## Building the metric properly

### Fix the question set

Pick 10–30 questions your customers actually ask: category recommendations, "alternatives to X," "is Y worth it," the works. This set is your panel — changing it mid-stream breaks your trend line, so draft it carefully and version it when you must change it.

### Ask all four engines, repeatedly

One engine is a sample size of one, and engines genuinely differ — Perplexity might cite you into every answer while Claude has never heard of you. Ask ChatGPT, Gemini, Perplexity, and Claude the same questions on a schedule. At Brandflare we call each run a sweep, and the differences between engines are routinely the most useful finding.

### Count presence, then quality

The core number is presence: the share of answers that name you. That's the [AI Visibility Score](https://brandflare.io/learn/glossary/ai-visibility-score) — averaged across engines over a trailing window (we use 7 days of sweeps, all four engines, 0–100). Layer quality on top: When named, are the facts right? Are you first-mentioned or an afterthought? Named enthusiastically or with caveats?

## Reading the number like an analyst

- **Level tells you standing.** A 40 means customers asking AI get told about you a bit under half the time. Your competitor at 70 is on shortlists you're not.
- **Trend tells you whether work is working.** Published the comparison content, fixed the stale sources? The next sweeps are your verdict. Retrieval-driven movement shows in weeks.
- **Per-engine spread tells you where the problem lives.** High on Gemini, low on Claude usually means your ranked content is fine but your broader footprint is thin — a memory problem wearing a metrics costume.

## The honest caveats

Engines are stochastic; the same question can produce slightly different answers on different days. That's why you average over a window instead of panicking over a single run. And share of answer measures visibility, not revenue — though a metric that counts "was my brand on the shortlist customers saw" sits about as close to revenue as brand measurement gets.

The fastest way to get a baseline is a [free audit](https://brandflare.io/audit) — real customer questions, all four engines, scored in about five minutes. The trend line starts whenever you do.