Measuring AI Share of Voice

Share of voice (SOV) is the single most useful number in AI search, because it’s inherently competitive. Your own visibility can rise while you still lose — if rivals rise faster. SOV answers the question that actually matters: of all the times AI names a brand for your buyers’ questions, how often is it you?

The definition

AI Share of Voice = your mentions ÷ (your mentions + all tracked competitors’ mentions), across a fixed set of buyer prompts.

If, across 20 test questions, brands get named 50 times total and you’re named 17 of them, your SOV is 34%. Simple — and powerful, because it turns a vague sense of “how are we doing?” into one comparable percentage.

How to measure it without fancy tools

You don’t need a live platform to do this well. You need a repeatable method:

Fix your prompt set. Write 20–40 real buyer questions and keep them the same every period. Consistency is what makes the trend meaningful.
Pick your competitors. Choose the 3–5 brands you’re genuinely up against.
Test in a neutral session. Use logged-out / incognito / temporary chat so your own history doesn’t skew results.
Sample each prompt. Run each question 2–3 times per engine and note which brands appear.
Tally mentions per brand, then apply the formula.

That’s it. A spreadsheet does the math. The discipline — same prompts, neutral sessions, repeated sampling — is what makes the number trustworthy.

Reading SOV well

Segment by engine. You might own the conversation on Perplexity but be invisible on Google AI Overviews. The average hides that; the breakdown reveals where to focus.
Watch the gap, not just your line. Closing from 20 points behind to 5 points behind is a win even if your absolute number barely moved.
Tie movement to action. When SOV jumps, note what changed — a new comparison page, a batch of reviews — so you can do more of what worked.

Why it beats a lone visibility score

A visibility score tells you how you’re doing. Share of voice tells you how you’re doing relative to the brands stealing your recommendations. In a channel where the assistant names only a few brands per answer, relative position is the whole point. SOV keeps you honest about whether you’re actually winning the conversation or just making noise in it.

Make it a habit

Measure SOV every cycle — quarterly, monthly, or weekly — with the same prompts and method. The absolute number matters less than the direction. Over a few cycles, the trend line becomes the clearest evidence you have that your AI-search work is paying off.