AI Visibility · 24 August 2026

By Eddie Eastwood, Founder

How to Measure Your Brand's Share of Voice in AI Answers

What is share of voice in AI, and how do you calculate it?

Share of voice in AI is the percentage of a defined set of AI-generated answers that name your brand, measured against how often a set of named competitors get named for the same questions. The formula: your brand's mentions, divided by the total mentions across your brand and your competitor set, across a fixed list of prompts and a fixed set of AI platforms, multiplied by 100.

Say you run 20 prompts across three platforms, 60 answers in total, and your brand appears in 9 of them while three named competitors together appear in 27. Your share of voice for that test is 9 divided by 36 (9 + 27), or 25%. That figure describes exactly one thing: how you and those three competitors split visibility on those 20 prompts, on those three platforms, on the day you ran the test. It is not a market-wide statistic, and it does not transfer to a different prompt list, a different competitor set, or a platform you didn't test. Building the prompt list and running the test itself is the subject of How to Check If AI Recommends Your Business; this page picks up from there, once you have the raw mention counts, and asks what the number actually means.

The number changes depending on which platform you ask

The single biggest source of confusion in AI share of voice is treating it as one figure when it behaves as several, one per platform, built from different evidence.

The platforms do not cite the same sources as each other. Wikipedia leads ChatGPT's citations while Reddit leads on both Google AI Overviews and Perplexity, and the sources beneath those leaders diverge further still; the figures behind that split, and what they mean for where a brand needs to be present, are set out in Who Can Get My Website to Show Up in ChatGPT?. For measurement, the consequence is the one that matters here: a share-of-voice figure carries the citation habits of whichever platform produced it, so the platform is part of the number, not context around it.

This reflects a real, structural difference in how the platforms are built. What Is AI Visibility? sets out why each platform runs its own retrieval process rather than sharing one index, and that difference is exactly why a share-of-voice number needs a platform label attached to it.

Does a share-of-voice number tell you what to change?

Not on its own, and this is where the metric is most often misread. A falling number can mean your own presence weakened, or that a competitor's strengthened while yours held, or that the platform changed which sources it draws on that month. The measurement identifies which of the three you are looking at only if you have held the prompt list and the competitor set steady, which is why the method section below matters more than the score. On what correlates with a higher score in the first place, the evidence is off-site rather than on-page, and it is covered with the figures in Who Can Get My Website to Show Up in ChatGPT?.

Reading your own number correctly

Once you have run a real prompt set through the method in How to Check If AI Recommends Your Business and have actual mention counts to work with, four things keep the resulting share-of-voice figure useful rather than misleading.

Report the platform alongside the number, always. "32% share of voice" means something different from "32% share of voice on ChatGPT specifically." Given how differently the platforms cite sources, the second version is the useful one, and it also tells you exactly where to focus next.

Treat the competitor set as part of the method. Swapping in a different set of three competitors changes the denominator and therefore the score, even if your own mention count hasn't moved. Naming the set is itself a decision: pick the businesses your buyers would actually weigh up against you, or the score ends up measuring a contest nobody but you is running. Fix the competitor set before you start tracking, and keep it fixed across every re-test so the number is comparable to your own past runs. If you do change it, for instance because a competitor exits the market or a new one becomes genuinely relevant, note the change alongside the new figure rather than letting the number appear to move on its own.

Decide what counts as a fixed denominator up front. A prompt list run once and never repeated cannot tell you whether a number moved because of a real shift in citation behaviour or because that day's answers happened to vary, which AI platforms do run to run even on an identical prompt. Running the same list on a fixed cadence, weekly or monthly rather than one-off, is what turns a single share-of-voice snapshot into a trend you can act on.

Don't chase one blended figure across all platforms. A single number that averages ChatGPT, Perplexity and Google AI Overviews together buries the platform-to-platform variation that is the most actionable part of the data. A brand at 40% on one platform and 4% on another is a different, more useful finding than a blended 22% that hides both.

FAQ

What is the formula for AI share of voice?

Your brand's mentions divided by the total mentions across your brand and a defined set of named competitors, for a fixed set of prompts and a fixed set of AI platforms, multiplied by 100. The number only means anything relative to the prompt list and competitor set you chose, so it is not comparable to another business's reported figure unless the method matches.

Can I measure share of voice on just one AI platform?

You can, but treat the result as that platform's figure. The platforms draw on noticeably different citation source pools by design. A single-platform figure describes that platform's behaviour.

Why does my share of voice look completely different on ChatGPT versus Google AI Overviews?

Because the two systems draw on different source pools by design. A brand well established on the sources one platform favours can be nearly invisible on a platform that favours different ones, which is a source-pool difference.

Can any tool give me one true share-of-voice score across every AI platform?

No single score reflects every platform equally, because the platforms do not draw on the same evidence. A tool can run the same prompts across several engines and report the results side by side, but a single blended number would flatten exactly the platform-to-platform variation that matters most for deciding where to focus.

Short version, restated

Share of voice in AI is a ratio you construct, and it changes by platform. Measure it per platform and keep your competitor set fixed, so that a change in the number means something happened rather than that you asked a different question.

The formula needs mention data before it can tell you anything, and that data starts with what the AI platforms currently return for your brand. Check whether AI platforms can find and cite your business →