AI share of voice is a brand's portion of the mentions or citations in AI answers to a defined set of questions, compared with every other brand or site that appears in the same answers. Where AI visibility asks whether you appear, share of voice asks how much of the conversation you hold against your competitors.
How AI share of voice is calculated
Fix four inputs first: the questions, the engines, the date window, and the market and language. Then run the questions repeatedly and count. There are three ways to count, and a report should say which one it uses.
| Measure | Counts | Divided by | Useful for |
|---|---|---|---|
| Mention share | Mentions of your brand | Mentions of all tracked brands | Comparing brands named in answers |
| Citation share | Citations of your domain | All citations in the same answers | Comparing sites used as sources |
| Answer share | Answers that name or cite you | All answers | Seeing how often you appear at all |
Repetition matters because answers change from run to run. Julius Schulte, Malte Bleeker and Philipp Kaufmann (2026) point out that in classical search a single query often gave a representative snapshot of where a brand stood against competitors, while AI answers vary across runs, prompts and time. Ronald Sielinski's 2026 study found that citation rankings were unstable across repeated samples and that many apparent gaps between domains fell within the measurement's noise. He argues that these figures should be reported with uncertainty estimates.
How to measure AI share of voice turns this into a procedure, and how to choose the prompts to track covers the first input.
Why AI share of voice matters
Raw counts cannot show whether you are gaining on competitors; share can. If your citations rise by 10 percent while all citations in the same answers rise by 50 percent, your count went up and your share fell. Share also makes engines comparable when they cite very different numbers of sources. In our scans (how we measured), Perplexity cited 20.2 third-party sources per answer and Gemini 3.5.
An example: Reddit on seven engines
| Engine | Citations to reddit.com | Third-party citations | Share |
|---|---|---|---|
| Google AI Mode | 94 | 1,814 | 5.2% |
| Google AI Overviews | 123 | 4,117 | 3.0% |
| Gemini | 55 | 1,910 | 2.9% |
| Perplexity | 300 | 11,131 | 2.7% |
| Grok | 5 | 2,481 | 0.2% |
| ChatGPT | 3 | 2,598 | 0.1% |
| Claude | 0 | 4,377 | 0.0% |
By raw count, Perplexity looks like Reddit's strongest engine, with 300 citations. By share, Google AI Mode leads: its 94 citations were 5.2 percent of everything it cited, almost twice Reddit's 2.7 percent in Perplexity, which spreads each answer across about 20 sources.
Common misconceptions
- "More citations means a bigger share." Not across engines, as the Reddit numbers show.
- "One run gives you your share." Answers vary between runs. Both 2026 studies above argue for repeated measurement, and Schulte, Bleeker and Kaufmann for treating visibility as a distribution rather than a single outcome.
- "An average across engines is enough." A blended figure hides the engines where you are strong and the ones where you are absent.
- "Share of voice and visibility are the same metric." Visibility counts your appearances. Share of voice sets them against everyone else's.