Share of Voice in AI: The Metric Every Brand Should Watch
When AI answers a question in your category, it names a few brands and skips the rest. Being one of the brands it mentions is good. But the number that really tells you how you are doing is not whether you appear, it is how much of the answer you own compared with your competitors. That is your share of voice, and in AI search it is the metric worth watching most closely.
Here is what it means, why it beats a raw visibility number for competitive decisions, and how to use it without falling into the common traps. We keep a short definition in the glossary; this is the strategic version.
What share of voice means in AI search
Share of voice is your portion of the brand mentions in AI answers for a given set of questions, measured against a defined set of competitors on the exact same prompts. If you ask twenty buyer questions across an engine and your brand is named in eight of the answers while a competitor is named in sixteen, that competitor has twice your share of voice on those questions.
The key words are "same prompts" and "against competitors." Share of voice is always relative. It is not how visible you are in the abstract, it is how visible you are compared with the others fighting for the same answer.
Why it beats "are we mentioned?"
A raw visibility score tells you whether AI knows you. Useful, but it misses the competitive picture. You can be mentioned more this month and still be losing, if your competitors are being mentioned even more. AI answers are a zero-sum space in a way that a page of ten blue links never was: the answer names a short list, and every slot a competitor takes is a slot you did not.
Share of voice captures that. It answers the question your CEO actually cares about, which is not "does AI know us?" but "when AI recommends someone in our category, how often is it us rather than them?"
How it is calculated
The mechanics are straightforward, and the fairness comes from holding the questions constant:
- You define the questions your buyers ask, branded and unbranded.
- You define the competitors you want to measure against.
- The same prompts run across each engine, and each answer is checked for who is mentioned.
- Your share is your mentions as a proportion of all the tracked brands' mentions, calculated per engine so the comparison stays clean.
Running the identical prompts for everyone is what makes the number trustworthy. Compare yourself on different questions from your competitors and the figure means nothing.
What a good or bad share of voice looks like
There is no universal target, because it depends entirely on your category and how many credible competitors there are. In a field of three serious players, a third is par and half is strong. In a crowded category, ten percent might make you the leader.
So do not fixate on the absolute number. Watch two things instead: your rank against the specific competitors that matter, and the trend over time. A share of voice climbing month over month is the signal that your work is paying off, whatever the starting point.
How to use it
Share of voice earns its keep when it drives decisions:
- Find where you are losing. Break it down by engine and you will often see you lead on one and trail badly on another. That tells you exactly where to focus.
- Pick the right competitor to worry about. The benchmarking makes it obvious which rival is taking the mentions you want, rather than guessing.
- Close the gap deliberately. Look at the answers where a competitor is named and you are not, see which sources those answers cite, and go earn presence in them. Share of voice tells you the score; the cited sources tell you why.
- Prove progress. A rising share of voice, tracked over time, is a clean way to show a client or a board that the investment is working.
Common mistakes
A few traps to avoid:
- Measuring on the wrong prompts. If your questions are not the ones your buyers actually ask, your share of voice is precise but irrelevant.
- Ignoring the per-engine picture. A single blended number hides the fact that you might be strong on Perplexity and invisible on ChatGPT. The engine breakdown is where the insight lives.
- Chasing the number instead of the trend. One reading is a snapshot, and AI answers vary between runs. The direction over time is what matters.
How to measure it
You cannot pull share of voice from your analytics, because the raw material is the AI answers themselves. Measuring it means running consistent prompts across engines, checking every answer for who is mentioned, and tracking the result over time against your competitors. That is exactly what Diploria does, and the Competitive Intelligence use case shows how it comes together.