Share of voice, defined
Share of voice measures how much of the AI conversation in your category belongs to you. You pick a set of prompts a real buyer would ask — "best CRM for small teams," "X vs Y," "tools for [job]" — run them across the major AI engines, and log which brands each answer names. Your share of voice is your mentions divided by all brand mentions across that prompt set.
It borrows directly from the traditional marketing metric, where share of voice meant your portion of total advertising spend or media coverage in a market. The AI version swaps "ad impressions" for "answer mentions." The logic is identical: the more of the category's attention you own, the more of its demand you're positioned to capture. The difference is where that attention now happens — inside a synthesized answer, not a ranked list of links.
3–5
Brands a typical AI answer names before it stops — which is exactly why a competitive share, not raw visibility, is the metric that matters.
Why a few mentions decide everything
AI answers don't return ten links — they synthesize one answer that names a short list, usually 3–5 brands, and frequently lead with a single recommendation. There is very little room. That scarcity is why share of voice, not raw AI visibility, is the metric that separates winners from also-rans.
How share of voice is calculated
The basic formula is simple: your brand's mentions divided by total mentions across every tracked brand, times 100.
SoV (%) = (your brand mentions ÷ total mentions across all tracked brands) × 100
A worked example: you track 1,000 AI answers across your prompt set, and your brand is named in 120 of them. That's a 12% share of voice. If a competitor shows up in 300, they hold 30% — more than double your slice of the conversation, and a clear signal of which prompts to go win back.
Vendor-grade measurement goes a step further and weights for **position**, because being named first is not the same as being buried at the end of an answer. A brand recommended first in every answer scores near the top; the same brand mentioned only as an afterthought scores far lower for the identical mention count. Some methods also weight by how often a prompt is actually asked, so popular questions count for more than obscure ones.
| Method | What it counts | Best for |
|---|---|---|
| Basic mention rate | Did the answer name you, yes or no | A fast first read on presence |
| Position-weighted | Named you + how prominently (first vs. buried) | Reflecting real recommendation strength |
| Volume-weighted | Position + how often the prompt is asked | Tying SoV to actual buyer demand |
Why it's the metric that predicts growth
Share of voice isn't just a scoreboard — decades of marketing research tie it to future market share. The principle, often called "excess share of voice," holds that when a brand's share of voice runs ahead of its share of market, the brand tends to grow; when it falls behind, the brand tends to decline. The brands AI engines name today are building the consideration that turns into sales tomorrow.
- 1
It's entirely earned
Unlike paid-media share of voice, you can't buy your way into AI recommendations. Engines synthesize answers from sources they trust, so your share reflects genuine authority — which makes it harder to fake and more durable when you win it.
- 2
It's competitive by construction
Because SoV sums to 100% across a category, it forces a head-to-head view. You don't just see whether you're visible — you see exactly who's taking the mentions you're not getting.
- 3
It's platform-specific
Share of voice varies sharply by engine. You might own 40% in ChatGPT but 15% in Perplexity. A single blended number averages those into a figure that's true nowhere — and the gaps are where the strategy lives.
How it relates to neighbouring terms
Share of voice sits next to a few terms it's easy to confuse. They're related, but they measure different things.
| Term | What it measures | Relationship to SoV |
|---|---|---|
| AI visibility | Whether AI engines mention you at all | SoV is visibility expressed as a competitive share |
| Mention rate | How often a single engine names you | A building block of SoV (your numerator) |
| AI citation tracking | Which exact sources an answer links to | The raw signal SoV is calculated from |
| Market share | Your slice of actual sales / revenue | SoV measures attention; market share measures results — SoV tends to lead it |
The most important distinction is the last one: share of voice measures *attention* inside AI answers, while market share measures *business results*. They correlate, and SoV usually moves first, but they are never the same number — treating them as one is the most common mistake people make with this metric.
Common misconceptions
- **SoV is not market share.** It measures presence in answers, not sales. The two correlate, and share of voice tends to lead, but they're different numbers.
- **SoV is not SEO rank or ad-impression share.** It's your prominence inside a generated answer — not a position in a list of blue links, and not paid impressions.
- **You can't buy it.** AI share of voice is fully earned. There's no ad slot to purchase inside the answer.
- **One global number is misleading.** SoV must be read per engine and per prompt set; figures diverge widely across ChatGPT, Perplexity, Gemini and AI Overviews.
- **Raw mention count isn't the whole story.** Being named first is worth far more than being buried — which is why serious measurement weights for position.
- **More keywords won't help.** Research on optimizing for AI answers found that adding statistics, quotations and authoritative citations lifts visibility, while old-school keyword stuffing barely moves it.
Want to see your share of voice across every AI engine? Run a free, one-page audit — it shows whether ChatGPT, Perplexity, Gemini and Claude name you, and how you stack up against the competitors named instead.
Run a free AI auditHow to measure and grow your share of voice
- 1**Define the competitive set.** List the brands a buyer would realistically weigh against you — those are the names you're splitting the answers with.
- 2**Pick category-defining prompts.** Cover discovery ("best tool for…"), comparison ("X vs Y") and use-case ("how do I…") questions a real buyer would ask.
- 3**Run them across every engine on a schedule.** Test ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews consistently — answers drift, so a one-off snapshot lies.
- 4**Log mentions, position and sentiment.** Record who's named, who's first, what's cited, and whether the mention is positive, neutral or negative.
- 5**Win the slices you're losing.** For prompts where a competitor takes your share, make your content more citable: lead with a direct answer, add statistics and quotes, and make sure AI crawlers can read you.
That measurement loop is exactly what SourceWatch runs for you. It tracks whether ChatGPT, Perplexity, Gemini and Claude cite your brand, turns those mentions into a share-of-voice figure against your named competitors — per engine, not a blurred average — and pairs it with the first-party AI-crawler and referral traffic actually reaching your site. So you can see your share moving, and see which prompts to fix next.