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

AI share of voice is the share of all brand mentions in AI answers, across a fixed set of questions, that goes to your brand. Ask ChatGPT, Perplexity, Gemini and Google AI Overviews a hundred buying-oriented questions, and if your brand appears among twenty of the hundred brands named, your AI share of voice is roughly twenty percent. This article explains what it measures exactly, how to calculate it step by step, how to line it up against your competitors, and where the limits of the measurement lie.

For the broader picture around visibility in AI search engines, read our guide to Generative Engine Optimization. Share of voice is one of the more concrete ways within that field to express your progress in numbers.

What is AI share of voice exactly?

AI share of voice measures how big your slice is of the brands that language models name when buyers in your category ask a question. It is the AI counterpart of the classic share-of-voice concept from advertising and SEO, where you set your share against the total market. The difference: here you are not counting ad impressions or search positions, but the number of times an AI answer drops your name relative to all the brands mentioned together.

A good measurement looks at three things at once. First frequency: how often you get mentioned at all. Then position: do you appear at the top of the answer or right at the bottom, because the first brand named weighs more heavily in the reader’s perception. And finally sentiment: are you described positively, neutrally or negatively. A brand that is mentioned often but lukewarmly has a different problem than a brand that surfaces rarely but always enthusiastically.

Important: share of voice is a relative measure. It says nothing about absolute volumes, but everything about your position within the competitive field. That is why it is so useful as a benchmark, and exactly why you have to set it up carefully.

How do you calculate AI share of voice?

The basic formula is simple: your number of brand mentions divided by the total number of mentions across all tracked brands, times one hundred. In practice you follow these steps.

1. Build a prompt set. Collect the questions your buyers really ask, not abstract keywords. Think of “best fleet management software for transport companies in Belgium” or “alternatives to [well-known player]”. Twenty to a hundred well-chosen questions give a far more stable picture than a handful.
2. Define your competitor set. Fix which brands you count. This is your denominator. Do not change that set between measurements afterwards, or your trends become meaningless.
3. Run the prompts in multiple engines. Test at minimum in the models your audience uses, such as ChatGPT, Perplexity, Gemini and Google AI Overviews. Every model draws on different sources, so your share differs strongly per engine.
4. Repeat every question several times. AI answers are not deterministic: the same question does not always produce the same answer. By running each prompt a few times and averaging, you filter out chance.
5. Count and divide. Count how often your brand is named and divide that by all brand mentions together. A worked example: if you are mentioned 40 times across all tests and all brands together 250 times, your AI share of voice is 40 divided by 250 times one hundred, or 16 percent.

If you would rather not do this by hand, specialised measurement tools exist. We list the options in our overview of AI visibility tools. We single one of them out in our Peec AI review on share of voice tracking for B2B. The calculation core stays the same formula regardless.

How do you compare your share with your competitors?

You compare your share with competitors by using exactly the same prompt set and engines for every brand you track, and then putting the outcomes side by side. Because share of voice adds up to one hundred percent across all brands, you see at a glance who dominates the conversation and where the room is.

Then break that comparison down, because the average often hides the most interesting insights:

• Per engine. You can be strong in Perplexity and barely register in ChatGPT, or the other way round. Models pick different sources, so a low share in one engine points to a targeted problem. We dig deeper into those differences in getting found in ChatGPT and getting found in Perplexity.
• Per question type. You may score well on questions about your exact category, but vanish as soon as someone asks about “alternatives” or “the cheapest option”. That tells you precisely which content is missing.
• Per stage of the buyer journey. Are you named in orienting questions but not in comparative ones? Then buyers see you early, but the model drops you at the decision moment.

That breakdown makes share of voice usable as a steering instrument instead of a loose number. Afterwards you know not only that you are lagging, but also where.

Why does your share differ so strongly per AI model?

Your share differs per model because every AI system draws on its own pool of sources and weighs those sources differently. One model leans heavily on current web results and citations, another more on what it picked up during training. As a result the same brand with the same questions can surface regularly in one model and stay virtually invisible in another.

That has a practical consequence: one average percentage across all engines hides too much. Always report your share per model, and repeat the measurement periodically, because models are updated continuously. How you influence whether a model picks up your information as trustworthy is closely tied to grounding, how AI determines what is true.

How do you increase your AI share of voice?

You increase your share by being confirmed consistently and widely as the answer to the questions your buyers ask. A high percentage is not bought with a single intervention; it is the result of a few things working together.

• Clear, citable content that unambiguously links your brand to a concrete category, audience and region. Models pull brand names from repeated, unequivocal associations.
• Independent mentions outside your own site. One voice about yourself weighs little; many consistent voices on authoritative sources make you hard to ignore. Why that counts for more than classic backlinks, you can read in brand mentions over backlinks.
• Consistent positioning across all your channels, so the model places you in the right context.

Share of voice is not the only gauge that counts, by the way. It fits into a broader set of indicators that we describe in the 5 core indicators of AI visibility. Improvement rarely happens overnight: count on a lead time of several months before a targeted programme has a visible effect on your share.

The short summary

AI share of voice is your slice of all the brand mentions AI models make on a fixed set of buying-oriented questions. You calculate it by dividing your mentions by the total and multiplying by one hundred, each time with the same prompt set, the same competitor set and multiple repetitions per question. The real value is not in the number itself, but in what the breakdown per engine, question type and buying stage shows you about where your content falls short.

At Customer Impact we deliberately do not steer on a pretty percentage to show off with, but on mentions that ultimately deliver leads and revenue. Want to know where your brand stands today in AI answers and where the biggest gain is? Take a look at our generative engine optimization (GEO) or schedule your free intake.