Share of voice (SOV) is the percentage of visibility a brand holds in a market compared with its competitors, on a given channel and over a given period. You calculate it by dividing the brand's metric by the total of the same metric across the market: advertising spend, ad impressions, estimated organic clicks, mentions, appearances in AI answers.
The formula is a division, so that's the easy part. The number changes a lot depending on what you put in the denominator: which competitors, which keywords, which channel. That's why, alongside the formulas, you'll find below a test we ran on the UK car market and the case of a B2B manufacturer that ranks ahead of Amazon on the keywords of its own niche.
What share of voice is
Share of voice started in traditional advertising as a brand's "advertising weight": its share of total advertising spend in the category. It's the definition Nielsen still uses: a brand's media spend expressed as a percentage of all media spend in the category, in that market, on that channel and at that time. Wikipedia adds that the weight is usually defined in terms of expenditure, ratings, pages or poster sites.
With digital, the idea has broadened. Today people talk about share of voice for organic search too, for Google ads, for mentions on social media and in the press and, for the past couple of years, for the answers of ChatGPT, Gemini and AI Overviews. The logic stays the same: how much of the attention available in a category ends up on you.
The general formula is this:
Share of voice = (brand metric / total of the metric in the market) × 100
If your company spends £50,000 on advertising in a quarter and the category as a whole spends £500,000, your advertising share of voice is 10%.
The formula, channel by channel
What changes from one channel to another is the metric. The table lines up the six versions you come across most often and shows where to find the data.
| Channel | Numerator | Denominator | Where to find the data |
|---|---|---|---|
| Advertising (traditional and digital media) | The brand's spend or GRPs | Total spend or GRPs in the category | Advertising monitoring data, for example Nielsen Ad Intel |
| Google ads | Your ads' impressions | Impressions you were eligible for | Impression share and Auction insights in Google Ads |
| Organic search | Estimated clicks on the keyword basket (volume × CTR for the position) | Total estimated clicks on the basket | Semrush, Ahrefs and other rank trackers |
| Social media and press | Brand mentions | Mentions of all brands in the category | Social listening and media monitoring tools |
| Share of search | Searches for the brand name | Searches for the names of all brands in the category | Google Trends, Google Ads Keyword Planner |
| AI answers | Prompts in which the brand appears | Prompts tracked | Repeated sampling on ChatGPT, Gemini, Perplexity |
Two notes on the table. Impression share in Google Ads is already a share of voice calculated by Google: the official documentation defines it as impressions received divided by total eligible impressions, and shows a dash when there's too little data. The Auction insights report adds the same metric for the competitors appearing in your auctions. If you run paid campaigns it's the number to look at first; we cover it in the guide to pay per click too.
The organic version, on the other hand, doesn't exist in Search Console, because Search Console tells you nothing about competitors. Rank tracking tools calculate it. In Semrush's Position Tracking guide, share of voice is weighted by each keyword's search volume and by estimated traffic: two sites with the same number of keywords on page one can have very different shares of voice if one is first on the big searches and the other on the small ones. Click estimates by position come from CTR curves: Sistrix's, calculated on more than 80 million keywords on mobile, gives 28.5% of clicks to position one, 15.7% to position two and 11% to position three.
Share of voice and market share: the ESOV rule
Marketers care about share of voice for one specific reason: it moves before sales do. The comparison that counts is with market share (share of market, SOM).
The difference between the two is called ESOV, excess share of voice. A brand with 10% of the market and 15% of the category's advertising spend has an ESOV of +5 points. One with 10% of the market and 6% of the spend is at −4.
The reference figure comes from Les Binet and Peter Field's work on the case archive of the IPA, the UK's Institute of Practitioners in Advertising: across 171 campaigns between 1980 and 2010, as Nielsen reports, a brand gained on average 0.5 points of market share for every 10 points of ESOV. It's an average across campaigns, not a law: it holds for categories with many buyers and repeat purchases, much less for a B2B company with forty clients.
The same Nielsen article gives an idea of how concentrated spend can be. In 2024, among US furniture retailers, Wayfair, Sleep Number, Rooms to Go and Ashley Furniture together held more than 50% of the category's spend, IKEA was at 7.8% and La-Z-Boy at 4.5% (Nielsen Ad Intel data, United States).
Share of search: the version you can calculate for nothing
Hardly anyone in the UK knows what their competitors spend, unless they buy advertising monitoring data. In 2020 Les Binet proposed an alternative that can be calculated with public data: share of search, meaning Google searches for your brand divided by searches for all the brands in the category.
In his presentation at the IPA's EffWorks Global in October 2020, Binet showed across three categories (cars, energy and mobile handsets) that share of search correlates with market share and moves ahead of it: for cars, the lead was up to a year. The reasoning is simple. Before buying a car, people search for the brand. If a brand's searches are growing today, some of those people will turn up at a dealership in the following months.
The tools cost nothing or close to it: Google Trends to compare up to five brands at a time over time, Google Ads Keyword Planner for absolute volumes. The limit is that it only measures the demand that goes through the brand name.
The test: share of search and market share for cars in the UK
We tried to repeat Binet's comparison on the UK market, with data anyone can check. For the ten best-selling brands in the UK in 2025 we took the monthly search volume for the brand name alone on Google UK (Google Ads data via DataForSEO, average from September 2025 to August 2026) and compared it with 2025 registrations from the SMMT, the Society of Motor Manufacturers and Traders. The UK market closed 2025 at 2,020,520 new cars, and these ten brands account for 54.1% of it, so share of search and share of registrations are calculated on the total for the ten. The first version of this test, on the Italian market, showed the same pattern: the premium German brands sat well above the line.
| Brand | Average monthly searches for the brand | Share of search (among the 10) | Share of 2025 registrations (among the 10) | Share of total 2025 market | Difference (points) |
|---|---|---|---|---|---|
| Volkswagen | 287,250 | 13.9% | 16.3% | 8.8% | −2.5 |
| BMW | 459,833 | 22.2% | 11.2% | 6.1% | +11.0 |
| Ford | 142,917 | 6.9% | 10.9% | 5.9% | −4.0 |
| Kia | 140,833 | 6.8% | 10.4% | 5.6% | −3.6 |
| Audi | 235,583 | 11.4% | 10.2% | 5.5% | +1.2 |
| Hyundai | 198,750 | 9.6% | 8.5% | 4.6% | +1.1 |
| Toyota | 208,500 | 10.1% | 8.2% | 4.5% | +1.8 |
| Nissan | 83,625 | 4.0% | 8.2% | 4.5% | −4.2 |
| Mercedes | 252,250 | 12.2% | 8.2% | 4.4% | +4.0 |
| MG | 60,708 | 2.9% | 7.8% | 4.2% | −4.9 |
Three things stand out.
First: across all ten brands the relationship is weak. The correlation between share of search and share of registrations is 0.46. Take out the three premium German brands, BMW, Mercedes and Audi, and on the remaining seven it rises to 0.71.
Second: the premium brands sit above the diagonal. BMW has 22.2% of the searches and 11.2% of the registrations, the largest gap in the table; Mercedes has 12.2% against 8.2%. Part of those searches is aspirational: people who look at a BMW and then buy something else, or who are searching for a used one. A high share of search on its own doesn't mean you'll gain share. It should be read against the brand's own history, not against its competitors'. Volkswagen, which in the Italian test was above the line, in the UK is below it: the UK's best-selling brand has only the third-largest share of search.
Third: the mass-market brands that sell more than they're searched for are MG (−4.9 points), Nissan (−4.2), Ford (−4.0) and Kia (−3.6). Our hypothesis is that their demand goes through model names more than the brand (the three best-selling models in the UK in 2025 were the Ford Puma, the Kia Sportage and the Nissan Qashqai, according to the SMMT figures published by CAR magazine), and that part of the sales go through channels (fleets, company cars, leasing) that don't start with a search. We didn't measure this in the test: it's the limit of a share of search built on the brand name alone, and the reason it's worth including the names of the main models or products.
Three more limits to declare. Google Ads volumes are rounded into bands (201,000, 246,000, 301,000 and so on), so differences under one percentage point shouldn't be read. Some brand names are ambiguous: "mg" also means milligrams, so MG's real share of search is even lower than the table shows, and "ford" picks up a surname. And the search period doesn't match the registrations year: Binet himself works on time series, because the signal is in the movement more than in the snapshot.
Organic share of voice: an Italian manufacturer ahead of Amazon
In organic search, share of voice answers a practical question: on the keywords that bring in customers, how much of the estimated clicks goes to your site and how much to your competitors?
The case we use most often is Macropix, an Italian manufacturer of LED walls and advertising displays. In July 2026, on the cluster of LED wall keywords measured with Semrush on Google Italy, the ranking was: macropix.it 25.55%, amazon.it 18.08%, hddsvision.it 10.17%, dominodisplay.com 4.33%, shop.agmultivision.it 2.28%, tsaled.com 1.74%.
Amazon has a domain authority no niche manufacturer will ever match. What it doesn't have is a page written by someone who makes LED walls, explaining the difference between an indoor and an outdoor model. On the same keywords Macropix went, between 2020 and 2025, from position 88 to position 2 for "monitor pubblicitario" (advertising display) and from nowhere to first for "ledwall" (LED wall, 4,400 searches a month in Italy). We've written about why specialisation matters more than size in the article on company size and SEO strategy.
The case also shows the right way to read the number. The 25.55% isn't valid for "all of Google": it's valid for a basket of keywords chosen by whoever does the analysis. If the basket also included "monitor" or "television", Macropix's share would very probably fall and Amazon's would rise. The basket should be built with keyword research, starting from the searches that bring in leads, and kept fixed over time; to understand how to read the individual positions that make it up, see the guide to keyword rankings.
Share of voice in AI answers
The newest version measures how often a brand appears in the answers of AI assistants. Siteimprove defines it as the percentage of tracked prompts in which the brand appears in the answer. The same article, updated in July 2026, flags the main problem: the same question asked twice can name different brands, so a single check tells you little and you need repeated sampling on the same prompts.
In practice that means choosing a fixed list of questions your customers really ask, repeating them at regular intervals on ChatGPT, Gemini and Perplexity and counting how often you and your competitors appear. The full method, with the Search Console AI report and GA4, is in the guide to AI visibility. To raise the number, what matters most is brand citations and mentions on third-party sources, and here the work overlaps with digital PR.
What a good share of voice is
There's no threshold that works for everyone. It depends on how many competitors there are and how large your market share is. Three figures give the order of magnitude.
On Amazon, an analysis by Pacvue of a consumer electronics category found 84 brands competing for the top ten keywords, and none held more than 10% of paid share of voice. In the same kind of analysis on Walmart, with only two brands active, the leader held more than 90%. In our UK car test the most searched brand, BMW, stops at 22.2% of the ten.
The criterion that holds up best is the ESOV one: a share of voice is good when it's higher than your market share and growing over time. 12% is an excellent number for a brand selling 6% of the category and a warning sign for one selling 20%. For a small business that doesn't know its market share, the useful comparison is with itself: the same basket, the same competitors, month after month.
How to measure it in practice
For most of the companies we work with, four steps repeated once a month are enough.
- Pick your real competitors, the ones you lose deals to, not the ones whose websites look like yours. Five to ten is enough.
- Build a basket of keywords from the searches that bring in leads or sales, 20 to 100 depending on the sector, and freeze it. If you change it every month the number is no longer comparable.
- Take the data from the right tool for each channel: Semrush or Ahrefs for organic, impression share and the Auction insights report in Google Ads for ads, Google Trends and Keyword Planner for share of search, a fixed sample of prompts for AI.
- Put it in the report next to the metrics that matter for revenue. In the guide to the SEO report we explain where it goes; in an SEO dashboard it's better to show the time series, not the month's value.
Share of voice on its own doesn't pay the bills. It's a leading indicator, and it only makes sense next to the SEO KPIs that measure leads and sales. If you want a basket built on your market and measured every month, it's part of the work we do in our SEO consultancy projects and, for assistants' answers, in our AI SEO service.
Frequently asked questions about share of voice
It's your slice of attention in a market: the percentage of advertising, searches, clicks or mentions that goes to your brand rather than to your competitors, measured on one channel and over a set period.
You divide the brand's metric by the total of the same metric in the market and multiply by 100. With 200 mentions out of 2,000 in the category, share of voice is 10%. The metric changes by channel: spend, impressions, estimated clicks, mentions, appearances in AI answers.
Share of voice measures visibility, market share measures sales. The gap between the two is called ESOV: according to Binet and Field's analysis of 171 campaigns, as reported by Nielsen, every 10 points of ESOV are worth on average 0.5 points of market share.
That you're the only one appearing in the basket being measured. It happens on a very small set of keywords, usually ones containing the brand name, or in a category with no active competitors. On a basket built to measure competition, 100% almost always means the basket is wrong.
It's a type of Amazon Ads campaign, not a measurement metric. With Sponsored Brands reserve share of voice, the advertiser secures the top-of-search placement for its own exact-match branded keywords, paying a fixed CPM agreed in advance. It's available in the UK.
It depends: it's good when it's higher than your market share and growing over time. In very crowded categories the values are low: on Amazon, Pacvue found 84 brands competing for the top ten keywords of a consumer electronics category, and none held more than 10%.
Something the industry rarely says
Share of voice is the easiest KPI to rig. You don't need to lie about any data: you just choose the denominator. Take out of the basket the keywords where you're weak, drop the competitor that beats you from the list, and the number goes up. That's why a share of voice without the list of keywords and competitors used to calculate it shouldn't be accepted in any report, ours included. Always ask for it with the basket attached and the date the basket was fixed.
