AI share of voice, explained
Share of voice is an old marketing idea: of all the attention in a category, how much is yours? In AI search it becomes a concrete measurement. Take a defined set of questions that real clients ask. Run them across the assistants that matter. Count how often each brand is mentioned, and how prominently. A brand mentioned in 12 of 40 answers, where its nearest competitor appears in 20, has a clear picture of where it stands.
The value is in the comparison and the trend, not in any single figure. AI answers vary between runs, between assistants and over time, so share of voice should come from prompt sampling: the same questions, run the same way, at regular intervals. Good reporting also separates mentions from recommendations and records whether the description was accurate, since being named with the wrong facts is not a win.
AI share of voice is useful for three decisions. It shows which questions and which assistants are weakest. It shows which competitors are being recommended instead and, through citations, which sources support them. And it gives a stable baseline for judging whether AEO or GEO work is having an effect.