Definition

What Is AI Share of Voice and How Is It Measured?

By Rahil Jain, founder of Chalk LabsUpdated October 20, 2026How we research

The short answer

AI share of voice is the portion of brand mentions in AI answers that go to you, compared with named competitors, across a fixed list of prompts. If your brand appears 12 times and all brands together appear 84 times, your share is about 14%. There is no industry standard formula, so state yours and keep it constant.

Share of voice = your mentions divided by all brand mentions across the same prompts and assistants. Visibility rate is a different number.
No standard definition exists. Vendors use different formulas, so scores from two tools are not comparable.
Use a fixed prompt set for at least a month, or the trend line breaks ¹.
One vendor study found one run a day gave nearly the same visibility figure as ten ³. A single manual run is not comparable.

What is AI share of voice?

AI share of voice measures how much of the conversation in AI answers belongs to your brand. You pick a list of questions buyers ask, run them in AI assistants, count which brands get named, and work out what fraction is you.

The idea comes from older marketing, where share of voice meant your share of ads or press coverage in a market. Here the market is the set of answers an assistant gives.

Be careful with the name. Semrush defines a related measure, AI visibility rate, as the share of your tracked prompts where your brand appears ¹. That counts prompts, not brands. Share of voice counts your slice of all brand mentions.

Imagine asking an AI assistant twenty questions your customers would ask, such as which company to hire. You write down every business it names. If your business is named in one answer out of every seven names given, your share of voice is about one in seven. That is the whole idea.

How is AI share of voice calculated?

Divide your brand mentions by the total mentions of every brand in your comparison set. Use the same prompts, the same assistants and the same time window for all brands.

Ahrefs' Brand Radar shows an AI share of voice figure, but its measurement guide gives no formula, a denominator or a prompt set ⁴. That is why you should write your own definition down and stick to it.

A mention here means the brand name appears in the answer text. A citation means the answer links to a page as a source ⁵. Track them separately, and decide up front which one your share of voice counts.

Worked example with illustrative numbers (not real data)

20 prompts x 3 assistants = 60 answers
Mentions: Brand A 18, Brand B 30, Brand C 24, You 12 -> 84 total
Share of voice = 12 / 84 = 14.3%
Visibility rate = answers that name you / all answers = 12 / 60 = 20%
Both numbers are valid. They answer different questions.

What did an AI tool search for when asked for the definition?

On 20 October 2026 a Claude agent with web search ran one search for this topic: AI share of voice definition. It did not search for a formula, a tool or a benchmark.

That is one run on one tool, so read it lightly. Our reading: a page should put the definition and formula in its first lines.

How do you set up a fair measurement?

Fairness comes from three fixed choices: the prompts, the competitors and the assistants. Change none of them mid-quarter.

Start with 20 to 30 prompts a real buyer would type, spread across awareness, comparison and purchase questions. Semrush suggests around 30 to start and says the prompt list must stay constant for at least a month or trend lines break ¹. Then list three to six competitors a buyer would actually consider.

Run each prompt in each assistant you care about. Semrush's tracker follows one platform per campaign, so tracking two means running the same prompts twice ¹. Record the date, the answer and which brands appear.

Write prompts as buyers phrase them, not as marketers do.
Keep a separate list of branded prompts and unbranded prompts. Report them separately.
Log the competitor set once and do not edit it.
Note any big site change on the date it happens.

How often should you run the prompts?

For tracking by software, once a day is usually enough, according to one vendor study. Profound compared two set-ups over 14 days in June 2026, each covering 753 prompts on seven platforms. Visibility was 78.7% with one run a day and 80.4% with ten, a gap of about 2 points ³.

Citation share was 10.24% at one run a day and 9.99% at ten ³. Profound sells daily tracking, and the study covers one two-week window, so treat the claim as a vendor view.

If you check by hand, the lesson is different. A single run of a single prompt proves nothing, because AI answers vary. Run each prompt on a set day each week and watch the trend over several weeks, not one reading.

What are the common mistakes?

The first mistake is to compare scores between tools. Each vendor picks its own prompts, engines and formula, so a high score in one tool and a low score in another say nothing about each other.

The second is to ignore tone. Seer Interactive argues that a visibility score counts a glowing mention and a lukewarm one the same, and that brands should also track how accurately and favourably they are described ². Add a simple tag for each mention: positive, neutral or negative.

The third is to chase the number. A share of voice that rises because you added easy prompts has proved nothing. Judge the change on the same prompts.

What tools can measure it?

You can start with a spreadsheet. Paid trackers automate the running and the counting, and several exist, including Semrush, Ahrefs Brand Radar and Profound. We make no ranking of them here.

Whichever you choose, check four things before paying: which assistants it covers, how many prompts you can track, whether it counts mentions or citations, and whether you can export raw answers. Raw answers let you audit the score yourself.

Our AI readiness score tool is a different thing: it checks whether your site is prepared to be found, not how often you appear. The full method is in how to measure AI share of voice.

What an AI assistant searches when it answers this

We gave the question below to AI search tools and recorded the web searches they ran behind the scenes. This is called query fan-out. Each phrase is a page you could publish to be found and cited.