AI answer share of voice is how often AI answers name or cite your brand compared with your competitors. Pick a fixed set of buyer questions, run them in each AI tool several times, and count. Vendors define the number differently, so write down your formula and compare your own trend over time, not other companies' figures.
Classic share of voice compares your brand with rivals across ads, rankings or social posts. AI answer share of voice does the same for the answers AI tools write. You ask a set of buyer questions, read the answers, and see how much of the attention goes to you.
There is no single standard. Ahrefs defines AI Share of Voice as the percentage of AI responses within your topic set that mention or cite your brand versus competitors. LLM Pulse defines it as your brand mentions as a share of all mentions for the brands you track, across a defined prompt set. One counts answers and the other counts mentions. Both are reasonable, and they give different numbers.
LLM Pulse describes three formulas in use. Pick one, name it in every report, and keep it fixed.
A related figure is AI visibility: the percentage of your tracked questions where your brand is mentioned at all. LLM Pulse treats visibility as an absolute number and share of voice as a relative one.
Because AI tools rarely give the same answer twice. In a SparkToro study published in January 2026, 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI results, 2,961 runs in all. The chance of getting the same list of brands twice was under 1 in 100 for ChatGPT and for Google's AI. Seeing two lists in the same order was rarer still, under 1 in 1,000.
The author, Rand Fishkin, still concluded that visibility percentage across many prompts, run multiple times, is a reasonable metric. He called exact ranking positions in AI answers unreliable. Two caveats: by their own account the authors are not professional researchers, and one works at an AI tracking startup.
Change the question list on a schedule, not every week, so the trend stays comparable.
We know of no public benchmark we would trust for startups. The number depends on how many brands the category has, so compare yourself with your own baseline and with two or three named rivals on the same questions. SparkToro's data points the same way: visibility in a narrow category with few options looks very different from a wide one.
No one can promise a result, and the levers are the same as for any AI visibility work.
On paid plans, Mergeflo runs a weekly scan that asks ChatGPT, Gemini and Claude your buyer questions and records whether your brand is mentioned, cited or recommended, plus Google results. It does not check Perplexity or Google AI Overviews. That shows your own presence for each question. For a share-of-voice figure, you also need to count which rivals appear when the same questions are run.
Around the scan, Mergeflo finds search gaps and buyer questions and drafts briefs and articles in your brand voice. You edit and approve them, and it publishes to Webflow when you press Publish. The Free plan includes a site scan and AI visibility check, with 250 credits once and no credit card (pricing).
There is no trusted public benchmark for startups. Compare yourself with your own baseline and with two or three named rivals on the same questions, measured over several weeks. The number depends on how many brands the category has.
No. LLM Pulse treats visibility as the percentage of tracked questions where your brand is mentioned at all, and share of voice as your mentions relative to all brand mentions. Visibility is absolute. Share of voice is relative.
Not as a main metric. SparkToro found that the same list of brands in the same order came up less than once in 1,000 runs, so a single position is mostly noise. Count how often you appear across many runs instead.
There is no proven minimum. SparkToro ran each prompt 60 to 100 times, and its author suggests asking at least that often to learn an AI tool's set of recommendations. That is more than a small team can do by hand. Start with a small list, repeat each question, measure for at least four weeks, and treat the result as a trend, not a precise score.
On paid plans, Mergeflo's weekly scan records whether your brand is mentioned, cited or recommended for your buyer questions in ChatGPT, Gemini and Claude. It does not check Perplexity or Google AI Overviews. For a share against competitors, you also need to count which rivals appear when the same questions are run.
