AI share of voice: definition, formula, and how to track it
The share of voice formula, the four channels marketers use it on, and how to apply it to what ChatGPT and Google's AI say.
Share of voice (SOV) is the percentage of a market's advertising, conversation or visibility that belongs to your brand, measured against the competitors you choose over a set period. You calculate it by dividing your brand's number (ad spend, impressions, mentions) by the total for every brand you track, then multiplying by 100. AI share of voice applies the same formula to answers from ChatGPT, Gemini, Perplexity and Google's AI features, and is sometimes called share of model.
Share of voice began as an advertising metric. Social listening and SEO teams borrowed it later, and AI tracking tools borrowed it again, because a raw mention count says nothing about how you stack up against the brands a buyer is also considering. For the AI version we suggest Hypervisible, the software we publish. It ranks your brand against the competitors you pick on the very same answers, which is the comparison the formula needs. If you'd like to see every option, our comparison of AI visibility tools lists what each one tracks and charges.
What is share of voice?
Share of voice is your brand's slice of a total you define yourself: a list of competitors, one channel, one period. Change any of the three and the number changes.
Nielsen sticks to the original meaning. A brand's share of voice, in its definition, is its media spending expressed as a percentage of all media spending in the category, for a given market, channel and time. Wikipedia describes it the same way, as an advertising measurement model that compares one company's media spend with the total for its product or category.
Marketing writers use a wider definition. Sprout Social calls it the share of the market conversation a brand owns compared with its competitors, and HubSpot presents it as a way to compare brand awareness against competitors channel by channel, using social mentions, paid clicks or website traffic.
Nielsen adds a useful limit: share of voice tells you whether a brand has the means to compete for attention, and says nothing about whether a campaign worked. Spend buys the chance to be noticed. The creative decides the rest.
How to calculate share of voice
Share of voice = your brand's measure ÷ the total measure for all the brands you track × 100. The measure can be dollars, impressions, mentions or clicks, as long as you use the same one for every brand.
An example, with invented numbers. A regional mattress company, call it Brand A, spent $3 million on advertising last year. Its three main competitors in the same states spent $8 million, $5 million and $4 million. The total for the four brands is $20 million, so Brand A's share of voice is 3 ÷ 20 × 100 = 15%.
The same company can have a very different number on another channel. If the four brands got 6,000 social mentions last quarter and Brand A got 1,500 of them, its social share of voice was 25%, well above its share of ad spend.
This is where a lot of share of voice reports go wrong. Put a national online mattress brand with $10 million of spend in those states into the set, and Brand A drops to 10% (3 ÷ 30). Brand A spent exactly what it spent before; the total grew by $10 million. Pick the competitor set before you look at any numbers, and don't quietly change it between reports.
Advertising share of voice
The classic version divides your media spend by the category's total spend. In paid search, Google Ads offers a close equivalent called impression share: the impressions your ads received divided by the total impressions they were eligible for. The auction insights report shows impression share for the other advertisers in the same auctions too, which tells you who you are sharing the page with.
Social share of voice
Social share of voice counts mentions of your brand on social platforms against competitors' mentions over the same period. Listening tools break it down by engagement, impressions, unique authors and sentiment. Read the tone before you celebrate, because a product recall can push a brand's share of voice up as fast as a launch.
SEO share of voice
In organic search, you compare how visible your site is against competitors' sites across a list of keywords, usually weighting each keyword by its search volume and the position you hold. Google Search Console only reports your own site, so competitor figures come from estimates made by SEO tools.
Share of voice in AI answers
In AI answers, share of voice is your brand's mentions divided by all mentions of the brands you track, across a fixed set of prompts sent to each AI engine. Where a results page listed ten links, an AI answer names a handful of companies, sometimes one, sometimes none.
Take a second invented case, in AI answers this time: an HVAC company in Phoenix (Brand A) tracks itself and three local competitors on 24 prompts that don't name any company, such as “same-day AC repair in Phoenix”. Each prompt goes to five AI engines, which makes 120 answers. Brand A is named in 30 of them, Brand B in 60, Brand C in 36 and Brand D in 24.
| Brand | Answers naming it | Mention rate | Share of voice |
|---|---|---|---|
| A | 30 | 25% | 20% |
| B | 60 | 50% | 40% |
| C | 36 | 30% | 24% |
| D | 24 | 20% | 16% |
| Total | 150 mentions | 100% |
Brand A's mention rate is 30 answers out of 120, or 25%. Its share of voice is 30 mentions out of the 150 that the four brands collected, or 20%. Think of mention rate as attendance and share of voice as airtime: the first tells you whether the engines bring you up at all, the second how much room you get next to the competition. If the engines start naming five companies per answer instead of three, Brand A can keep its mention rate and still lose share of voice. Our AI visibility explainer goes further into mention rate and what a good score looks like.
The formula is blunt. Being the fifth name in a list scores the same as being the first, and “avoid this company” scores the same as “highly rated”. Read it next to average position and sentiment. Some trackers weight each mention by its place in the answer; whatever rule you pick, keep it the same every time you measure.
Share of model: the same idea, different name
Share of model is the name a Harvard Business Review article gave in June 2025 to a brand's presence in the answers of large language models. In practice it is AI share of voice, with a couple of extra measures.
The article, published June 4, 2025, was written by David Dubois, a marketing professor at INSEAD, and John Dawson and Akansh Jaiswal of the agency Jellyfish. They measure share of model through three lenses: mention rate, a human-AI awareness gap (how well people know a brand in surveys compared with how often the models bring it up), and the sentiment the models express about the brand and its category.
Their example is the best argument for splitting results by engine. In the Italian laundry detergent market, Chanteclair had a 19% share on Perplexity and did not appear at all on Llama. Ariel ranged from almost 24% on Llama to less than 1% on Gemini. An average across engines would have hidden both gaps.
Share of voice vs. market share
Market share counts sales; share of voice counts presence. Nielsen's rule of thumb is that a brand whose share of voice exceeds its market share is likely to grow over the long term, while one that stays below tends to decline.
The gap between the two is called excess share of voice, or ESOV. Nielsen points to work by Les Binet and Peter Field for the IPA, the UK's advertising agency association: across 171 campaigns from 1980 to 2010, a brand's market share typically rose by about half a point for every 10 points of ESOV. Nielsen notes that the effect varies by channel, brand size and category. In the mattress example, a 15% share of voice against a 10% market share would be an ESOV of 5 points.
In 2020, Binet proposed a related measure, share of search: Google searches for a brand's name divided by searches for all brands in its category, taken free from Google Trends. According to the IPA, it correlated with market share in cars, energy and mobile handsets, and ran ahead of it by up to a year for cars.
As far as we know, nobody has published that kind of evidence for AI share of voice yet, and a precise ratio would be a guess. Our view is that the link will show first where buyers ask for a shortlist: a homeowner who gets three names from ChatGPT has little reason to go looking for a fourth.
How to track it every week
Fix the prompts, the competitors and the engines, take a reading on the same schedule, and watch the change from one reading to the next more than the level itself.
- Choose the competitor set once. Start with the companies your sales team actually loses deals to. If the engines keep recommending a business you've never heard of, it belongs on the list too; log the date you added it so the trend line still makes sense. Our guide to competitive intelligence in AI search shows how to spot them.
- Use prompts that don't name you. A prompt that includes your company name will produce an answer about your company, which tells you nothing about the competition. Keep branded prompts for a separate check of what the AI says about you.
- Ask each engine, more than once. The same prompt on Monday and on Tuesday can bring back different companies. Several runs per prompt keep one odd answer from skewing the week.
- Count one mention per brand per answer. An answer that praises a competitor in three separate paragraphs is still one mention.
- Report by engine and in total. A blended number would have hidden Chanteclair's zero on Llama.
- Look at the sources. The pages an AI cites when it names your competitors tell you where to work next.
One reading a week on unchanged prompts gives you four or five points a month, plenty for a trend and a monthly report. Daily data is worth paying for during a launch or a bad news cycle. An AI rank tracker automates the routine, and agencies that report to clients can borrow the format from our guide to AI visibility reporting for agencies.
Tracking AI share of voice with Hypervisible
Hypervisible, which publishes this guide, takes the weekly reading for you, and it was built around the measures this article says to read together.
Each Monday it repeats last week's questions in five engines: ChatGPT, Perplexity, Gemini, and Google's AI Mode and AI Overviews. Out come your citation rate, your rank among the competitors you follow on those same answers, your average position inside each answer and the tone the engines take about you. Those last two cover the blind spots of the raw formula, where fifth place counts like first and a warning counts like praise. Each figure shows its change in points since the previous Monday, so the trend sits on the page.
Google's two AI engines come with every plan, which matters when your buyers start their search on Google. The tool also keeps markets apart, reading four of them (US, UK, France, Germany) and asking in English, French or German as each requires, so a brand selling in Phoenix and in Manchester doesn't blend two competitor sets into one figure. The sites the engines cite are listed as well, which handles step 6 above. Behind it is NEXUS GEO, an agency whose only trade is GEO; data from Google and the AI engines feed each Monday reading.
What it leaves out: the reading is weekly, and Claude, Copilot and Llama aren't measured. If you haven't measured anything yet, try the free AI visibility checker first. It sends ten questions to the same five engines (half on your industry with no brand named, half naming you) and has the 50 answers read in 5 to 10 minutes. Nobody asks for a credit card. Prices for the paid plans are on our pricing page.
The free analysis asks 10 questions to five AI engines (ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode), reads the 50 answers in 5 to 10 minutes and keeps your dashboard in your account.
Frequently asked questions
How do you calculate share of voice?
Divide your brand's measure by the total for all the brands you track, then multiply by 100. The measure depends on the channel: ad spend, ad impressions, social mentions or mentions in AI answers. In a made-up example, a brand spending $3 million in a category where four brands spend $20 million has a 15% share of voice. The result depends on which competitors you include, so keep the same list every time you measure.
What does 50% share of voice mean?
A 50% share of voice means your brand accounts for half of the spend, impressions or mentions counted across the brands you track in that period. How good that is depends on how many competitors you track. Against one rival, 50% is a tie. Against nine, it is five times the 10% average. In AI answers, it means one of every two brand mentions in your tracked set is yours.
What is share of model?
Share of model is the term a June 2025 Harvard Business Review article used for a brand's presence in the answers of AI models such as ChatGPT, Gemini or Llama. The authors measure it with mention rate, the gap between how well people and AI models know the brand, and the sentiment the models express about it. The core calculation is the same as share of voice: your mentions against the total for your category.
How do you measure share of voice in ChatGPT?
Hypervisible, the software we publish, does it every Monday on the same questions and ranks you against your competitors in ChatGPT, Gemini, Perplexity and Google's AI answers, with the change in points. By hand, write a fixed list of prompts about your category that don't name your brand, ask ChatGPT each one several times, and note every brand named. Divide your mentions by the total mentions of the brands you track and multiply by 100. Repeat on a schedule, since answers vary between sessions.
What is a good share of voice?
There's no universal benchmark, because the number depends on how many competitors you track and which channel you measure. A useful reference is your market share. Nielsen's rule of thumb holds that a brand whose share of voice is above its market share tends to grow, and one below it tends to shrink. For AI answers, compare yourself with the brand the engines name most often, and watch the trend from week to week.
What is the difference between share of voice and market share?
Market share is your share of category sales. Share of voice is your share of the category's advertising, conversation or visibility. The two are linked: in analysis by Les Binet and Peter Field covering 171 campaigns, cited by Nielsen, market share typically rose by about half a point for every 10 points by which share of voice exceeded it. That gap is called excess share of voice, or ESOV.
Sources
- Share of voice, Wikipedia, read October 22, 2026
- Need to know: What is share of voice?, Nielsen, March 2025, read October 22, 2026
- Share of voice: What it is and how to measure it, Ronnie Gomez, Sprout Social, March 11, 2026, read October 22, 2026
- The Beginner's Guide to Share of Voice, Rebecca Riserbato, HubSpot, updated February 6, 2025, read October 22, 2026
- About impression share, Google Ads Help, read October 22, 2026
- Use auction insights to compare performance, Google Ads Help, read October 22, 2026
- Forget What You Know About Search. Optimize Your Brand for LLMs, David Dubois, John Dawson and Akansh Jaiswal, Harvard Business Review, June 4, 2025, read October 22, 2026
- Same article, full text distributed by The New York Times Syndicate, American Association for Physician Leadership, read October 22, 2026
- Binet presents fast, cheap, predictive Share of Search metric to EffWorks Global 2020 Conference, IPA, October 14, 2020, read October 22, 2026