AI Share of Voice: How to Measure It Without Guesswork

Key takeaways

AI share of voice (AI SoV) is the percentage of AI answers in your category that mention or cite your brand, compared with competitors.
Tools count it differently: by mentions, by answers or by citations. Pick one formula and never mix them in the same report.
AI answers are unstable. In a SparkToro study there was less than a 1 in 100 chance of getting the same brand list twice, so you need many prompts and repeated runs.
You can measure it for free with a spreadsheet; Bing and Merchant Center also give partial official versions.

2,961AI runs in SparkToro's study on how often brand recommendations change

4.5 weeksaverage time a source stays cited before it rotates out (Scrunch)

3common ways to count it: mentions, answers or citations

What is AI share of voice?

AI share of voice is the percentage of AI-generated answers about your category that name or cite your brand, measured against the competitors you track. It adapts the classic marketing metric to ChatGPT, Gemini, Perplexity and Google's AI features. It shows whether AI assistants put you in the conversation when buyers ask for a recommendation.

Ahrefs defines it on its Brand Radar page as the percentage of AI responses within your topic set that mention or cite your brand versus competitors. It is useful because a single answer tells you almost nothing; a share across many questions tells you where you stand. It sits next to the other GEO measures explained in the GEO hub.

How do you calculate AI share of voice?

Divide the AI mentions of your brand by the total mentions of all tracked brands, then multiply by 100. That is Semrush's formula. Ahrefs counts answers instead: answers that mention you divided by answers that mention any tracked brand. Bing's Citation Share counts citations of your site out of all citations for a query.

Version	Formula	Source	Best for
Mention share	Your mentions ÷ mentions of all tracked brands × 100	Semrush	Categories where answers list many brands
Answer share	Answers that mention you ÷ answers that mention any tracked brand × 100	Ahrefs	Small businesses; easy to count by hand
Citation share	Citations of your site ÷ all citations for the same query × 100	Bing Webmaster Tools	Content sites that want links, not only mentions

Both Semrush and Ahrefs add weighting in their tools (Semrush also considers how high a brand appears in the answer; Ahrefs weights prompts by estimated traffic). That makes their scores useful for trends but not comparable with each other, or with your own spreadsheet. Choose one version and keep it.

A worked example

This is an illustrative example, not client data. A dental clinic tracks 20 buyer questions ("best dentist for implants in Medellín", "cuánto cuesta un implante dental en Medellín") in two assistants, three runs each: 120 answers. In 80 of them at least one tracked clinic is named.

Brand	Answers that mention it	Answer share (÷ 80)
Your clinic	18	22.5%
Competitor A	46	57.5%
Competitor B	31	38.8%
Competitor C	12	15.0%

The shares add up to more than 100% because one answer can name several clinics; that is normal for answer share. What matters is the gap: competitor A is named more than twice as often. The next step is to read the sources the AI cited in those 46 answers. They tell you which directories, review sites or articles to work on. The GEO strategy guide explains how to turn that list into a plan.

How many prompts and runs do you need?

More than most people use. AI answers change from run to run, so a handful of prompts produces a number that jumps every month. Use at least 20 buyer questions, run each one two or three times per assistant, and compare three months before you call it a trend. Keep the same questions, language and settings every time.

The SparkToro study behind this advice had 600 volunteers run 12 prompts 2,961 times in ChatGPT, Claude and Google's AI. It found less than a 1 in 100 chance that ChatGPT or Google's AI gives the same list of brands twice in 100 runs, and about 1 in 1,000 for the same order. Its conclusion: a visibility percentage across dozens to hundreds of prompts run multiple times is a reasonable metric; a "ranking position in AI" is not. Sources also rotate: Scrunch measured an average of 4.5 weeks before a cited source is replaced, and 3.4 weeks in ChatGPT. Ahrefs gives the same advice in its AI share of voice guide: trust trends that persist across several reporting periods, not one-off spikes.

Can you measure AI share of voice for free?

Yes. Put your buyer questions in a spreadsheet, run them in each assistant on the same day each month, and mark which brands each answer names and which sources it cites. Bing Webmaster Tools gives a free Citation Share for Copilot, and Merchant Center shows AI share of voice for stores in five English-speaking countries.

Bing's Citation Share does not show competitor names, only your share. Google's Merchant Center AI insights compare you with competitors you define, for English shopping queries on AI Mode and AI Overviews in Australia, Canada, India, New Zealand and the US. Everyone else, including most businesses in Latin America, needs the spreadsheet. Run your questions in Spanish if your buyers ask in Spanish; the answers and the sources are often different. For paid trackers that automate this, see the GEO tools guide.

What is a good AI share of voice?

There is no universal benchmark. A good AI share of voice is one that beats your own last quarter and closes the gap with the competitor buyers compare you with. A local business named in a third of category answers in its city is in a strong position; a national brand in a crowded category may be happy with less.

Connect the number to money: watch GA4's AI Assistant channel and the leads or sales that come from it. A rising share with flat leads usually means you are named for questions that do not buy. My GEO plans report AI share of voice every month with the sources behind it, and the home page explains how I work.

Frequently asked questions

Is AI share of voice the same as classic share of voice?

Same idea, different place. Classic share of voice compares your visibility or spend with the market; AI share of voice compares how often AI answers name you versus competitors.

Should I count mentions or citations?

Mentions show whether AI recommends you; citations show whether it links to you. Track both if you can, but report them separately.

Why does my AI share of voice change every week?

Because AI answers vary between runs and cited sources rotate every few weeks. Use more prompts and repeated runs, and compare monthly trends.

Which AI assistants should I include?

The ones your buyers use. For most businesses that means ChatGPT and Google's AI features first, then Gemini, Perplexity or Copilot depending on the audience.

Does a higher AI share of voice mean more sales?

Not automatically. Check that the questions you track are buying questions, and connect the metric to AI referral visits and leads.

About the author

Juan José Giraldo is an SEO consultant in Medellín, Colombia, with 5+ years of experience. He works on technical and local SEO for 600+ US car dealerships at Reunion Marketing, an Inc. 5000 agency, and supervises the SEO of 18 US online stores with more than 3 million products at Prontomowers. Read more · LinkedIn.

AI share of voice: how to measure it without fooling yourself.