Plain-English definition
How often AI assistants mention your brand when buyers ask category, comparison, or purchase questions.
The percentage of AI-assistant answers about your market that name your business. The honest measure of whether ChatGPT, Perplexity, and Claude know who you are when buyers ask — measured by reading what the AIs actually say, not by inferring it from SEO proxies.
Formula
AI SoV
(answers naming you ÷ total answers) × 100
Measures
Observed mentions
Direct output measurement, not indirect ranking signals.
Use it for
Trend tracking
Directional monitoring — track whether the line goes up.
A category-defining metric needs to fit on a single page. Here it is: what it is, how it is calculated, what it is good for, and where its limits sit. Everything that follows is elaboration.
Plain-English definition
How often AI assistants mention your brand when buyers ask category, comparison, or purchase questions.
Formula
(answers naming you ÷ total answers) × 100
Best used as
Monitoring direction over time — not as a single absolute reading. AI SoV is a trend signal, not a scorecard.
Core caveat
You can observe whether you were named. You cannot definitively observe why the model named you. Models are opaque.
Three ingredients: a representative prompt set, a defined provider list, and a deterministic detection rule. Each prompt is asked of each provider; each cell is scored for a mention; the percentage is the share.
Five to twelve natural-language questions a real customer would ask — covering commercial intent (“best X”), informational intent (“how does X work”), and comparison intent (“X vs Y”). The mix matters because each intent type tests a different assistant behavior.
Perplexity (live-web retrieval), ChatGPT (parametric training data), and Claude (parametric training data). Three providers because they represent the actual surface area of customer use today — and because parametric and live-search assistants behave differently enough that a single provider is misleading.
For each (prompt × provider) cell, detect whether the business name appears in the answer. Sum the appearances, divide by total cells, multiply by 100. The output is a single number on a 0–100 scale and a per-provider breakdown.
A blended AI Share of Voice is a useful headline, but it averages over real differences. The same 42% can mean three very different positions — and the right move depends on which one you are actually in.
The blended share across the full prompt set and all providers. Useful as a single tracking line over time — good for telling whether visibility is improving, less good for telling you why.
The split across Perplexity, ChatGPT, and Claude. Often the most strategic view: live-web assistants react to recent press and social proof; parametric assistants react to long-tail authority signals. Where you are weak tells you where to invest.
The split across commercial, informational, and comparison prompts. Tells you which buying moment you actually win. Winning “best X” is worth more than winning “how does X work” — but only if you sell X.
Treat these as editorial guidance, not scientific absolutes. AI SoV is directional — a 35 today versus a 35 next quarter is the comparison that matters more than the absolute band.
0–10
Invisible
Largely absent from relevant AI answers. The brand is not in the consideration set buyers are exposed to. Building from zero.
10–30
Occasional
Appears intermittently, not consistently enough to shape category perception. Foundation is real but uneven across prompts and providers.
30–60
Competitive
Regularly named in answer sets. Likely entering shortlists. The 42% Portland-roaster example below sits here.
60+
Category-leading
Named unusually often. Universal mention is rarely realistic in competitive categories — competitors also appear in most answers, so even leaders rarely break 80.
✓ What it measures
Whether AI assistants actually name your business when customers ask about your market. This is empirical — the AI's output is read directly, not inferred from web signals. Direct measurement beats SEO-style inference because it answers the question the buyer actually faces.
✕ What it does NOT measure
The causal reason an AI named you. AI ranking logic is opaque, and the major providers don't publish how recommendation is decided. Models also update, which can shift what they say about the same brand without anything you did changing. Track the trend, not any single reading.
Twelve customer prompts (“best specialty coffee roasters in Portland,” “where to buy single-origin beans online,” “Portland coffee subscription vs grocery store”) asked of three AI assistants. Thirty-six total cells. Fifteen mentions. The math:
The math
Total mentions: 15 of 36 cells.
AI Share of Voice: 42%.
Strategic readout
Origins
Share of Voice originated in 1980s advertising as a brand's presence in paid media relative to competitors — typically calculated as (your impressions ÷ total category impressions) × 100. AI Share of Voice extends the concept to AI-assistant answers as the new discovery surface: instead of counting paid impressions, it counts unpaid AI recommendations. Closer in spirit to organic search visibility than paid media. Wadsworth proposed the term in 2026 as the industry began needing a standardized name for what was already being measured ad-hoc.
Measure yours
Free. About fifteen seconds. You'll see whether AI assistants name you for the questions your customers actually ask — with the per-provider split, so you know where to push.
Run the free check →Cite this entry