In AI visibility, share of voice is the percentage of relevant AI answers that mention a brand, across a defined set of prompts, models, and time.
Classic share of voice measured advertising weight; the AI version measures answer weight. If ChatGPT is asked 100 buying questions in a category and names a brand in 40 of the answers, that brand holds a 40% share of voice on ChatGPT for that period.
Share of voice and rank answer different questions. Rank tells you where you stand when you are named; share of voice tells you how often you are named at all. A brand can rank #2 with a thin share of voice. It is recommended prominently, but rarely.
Why it matters now
Rank answers "where do I stand when I am named?" Share of voice answers a blunter question: "how often am I named at all?" In AI answers that second question decides reach, because a brand can hold a strong rank on the rare occasions it appears and still be missing from most of the answers buyers actually see.
It matters now because AI answers are narrow. A results page has room for many brands; an answer has room for a few. Share of voice measures how much of that scarce space a brand occupies across the category's questions, which is the closest thing to reach in a world with no impressions to count.
How WhereDoIRank measures it
We hold a fixed set of the category's buying questions and ask them to ChatGPT, Claude, Gemini, and Perplexity on a recurring schedule. Share of voice is the fraction of those recorded answers that name the brand, and it is computed per model, so you can see a brand that saturates one model's answers while barely surfacing on another.
Because the prompt set and schedule are constant, the figure is comparable close to close. A rising share of voice means the brand is being named in more of the same answers, not that we changed the questions underneath it.
A worked example
Suppose ChatGPT is asked a category's buying questions and names Brand X in a large share of the answers but usually late in the list, while it names Brand Y rarely but always first. Brand Y might hold the better average rank; Brand X holds the larger share of voice. They are winning different things: Y is prominent when present, X is present far more often.
Read together, the two numbers set the strategy. A high rank with a thin share of voice says "get named more"; a high share of voice with a weak rank says "get named earlier." The pair points to different work.
Common misconceptions
Share of voice is not the same as rank, and the two can move in opposite directions. It is also not a share of a fixed pie in the old advertising sense; here it is the share of relevant answers, over a defined set of prompts, models, and time, so the denominator is the questions asked, not a media budget. And a high share of voice is not automatically good news: if the mentions are caveated or dismissive, the brand is named often and framed poorly, which is why sentiment is read alongside it.
How it connects
Share of voice is built from brand mentions, the atomic events it counts, and it feeds the AI Visibility Score alongside position and consistency. It depends on prompt tracking to define the fixed question set it is measured over. Rank and share of voice together give the full picture: how prominent you are when named, and how often you are named at all.