Skip to main content

Share of voice

Share of voice, commonly abbreviated to SOV, is the portion of attention within a topic that goes to your brand rather than someone else's. It is a ratio, not a count: not how often you are talked about, but how often that happens set against every time the topic is discussed at all.

Published · checked

Who it is for

For anyone comparing themselves to others and having to back it up. Forty mentions in a month is a lot in a quiet market and very little in a busy one; only as a ratio does it say anything.

In practice the question comes from someone defending a budget or accounting for a campaign. They need to know whether the picture shifted against the competition, not against last month.

What we measure, and where

7
AI assistants queried
30
day window, with the previous period beside it
2
counts per answer: named and cited

We calculate a share in the one place where the denominator genuinely exists: in the answers AI assistants give to questions without your brand name in them. There you can count how many providers are named and how often you are among them. For news media and social channels we do not, because the denominator would be a guess.

Per answer we record the full answer, the cited sources with their URLs, every brand named in order of prominence, the sentiment, and whether the answer is genuinely about your organisation or about a namesake. That last field exists because it regularly goes wrong.

How many questions run and how often they are asked depends on the plan and on what you want to follow.

Visibility, our headline figure

visibility = answers naming you / all measured answers

  • answers naming you: the answers within the window in which your organisation appears
  • all measured answers: every usable answer to the questions without a brand name, across all platforms queried
  • window: thirty days
  • This is a presence rate: it says in how many answers you appear, not what portion of all brands named is you.

Seven questions across eight platforms gives 56 answers per round, or 224 over four rounds. Appear in 54 of them and visibility is 54 / 224 = 24 percent.

Answer · "which provider for reputation monitoring"
  • 1. Competitor Anamed, source: own website
  • 2. Acmenamed, source: trade publicationyou
  • 3. Competitor Bnamed, no source
  • Sources cited: 9of which 2 from acme.com
Per answer we record the brands named in order, plus which sources were cited and which of those are yours.
  • ChatGPT
  • Google Gemini
  • Perplexity
  • Grok
  • Google AI Mode
  • Google AI Overviews
  • Claude
  • Microsoft Copilot

You are in the answer or you are not.

Being named and being cited are two things

An assistant can name you without using your site, and use your site without naming you. So we count them separately.

Named and cited is the strongest position: the model knows you and uses your own words. Named but not cited means someone else supplies the story, and the question becomes whose page that is. Cited but not named is the silent contribution: your text feeds the answer while another name sits on top of it.

Beside the share sits prominence: where you are named and how many brands appear in total. That order comes from a judgement by a language model, which lists the providers named with the most central brand first. Platform names, source citations and place names do not count, and the list stops at fifteen.

Prominence

prominence = your position in the list, with the number of brands named beside it

  • Not a fraction. We do not divide those two, because that removes exactly the difference you want to see: first of four and third of twelve produce the same result, while the first is far stronger.
  • position: 1 if you are discussed first, 2 if second, and so on
  • number of brands named: how many providers appear in that answer, capped at fifteen

Named second in an answer with five providers is recorded as 2 of 5. That is a different thing from 2 of 12.

Prominence over a period

average position = sum of your positions / answers naming you

average brands named = sum of brands named / answers naming you

  • Both averages run over the same answers, the ones that name you. Where you do not appear there is no position to average, and the second figure would not be comparable either.
  • The two belong together: position 1.4 in an answer naming 4.8 brands on average is strong, while position 1.4 among two names is simply the lower half.
  • And the share belongs with them, otherwise you do not know in how many answers any of this applied.

The full reading is one sentence: you are named at position 1.4 on average, out of 4.8 brands named on average, in the 40 percent of answers where you appear.

Citation share

citation share = sources on your domain / all cited sources

  • sources on your domain: the cited sources that sit on your own domains
  • all cited sources: everything the assistant cites for that answer
  • Calculated per answer, then averaged across the window.

An assistant citing twelve sources of which four are yours gives a citation share of 4 / 12 = 33 percent.

How to calculate share of voice

The arithmetic is not the hard part; the three decisions before it are.

  1. 01

    Define the topic. What are you comparing yourself within: the whole category, one product, one region. Anything outside it counts on neither side of the fraction.

  2. 02

    Decide which brands sit in the denominator. This is where it usually goes wrong. Do you count only the competitors you named yourself, or everything that gets named? That difference moves the result more than any campaign will. We count every brand mentioned, because otherwise you are measuring your own assumptions. A newcomer you had not noticed simply lowers your share, and that is exactly what you want to know.

  3. 03

    Pick a window and keep it. A window that shifts with every report makes any comparison with the previous measurement useless.

Share of mentions, the classic share of voice

share = your mentions / all brand mentions in the same answers

  • This is the figure the term share of voice has always denoted: your portion of the attention, not your presence.
  • It comes out of the same measurement. For every answer we record each brand named with its position, plus whether it is your organisation, a competitor you track, or a provider you had not yet noticed.
  • One limit: we record at most fifteen brands per answer. For an answer naming more than that the denominator is truncated. It is rare, and it is stated here because it affects your figure.
  • The two diverge widely and that is not an error. In an answer listing twelve providers you are present, but you are also only a twelfth of the attention.

Appearing in four of ten answers that together name 45 brands gives visibility of 40 percent and a share of mentions of 4 / 45 = 9 percent.

What SOV stands for, and the difference with share of market

SOV is the common abbreviation of share of voice, and you will meet it in reports more often than the words themselves. Outside this field the same three letters are used for other things, so mind the context.

Share of market is your portion of revenue, share of voice your portion of attention. The gap between them is called excess share of voice, or ESOV. For a small brand in a crowded category that is an uncomfortable sum: to grow you have to be more visible than your size justifies.

What the research shows, and where the limit is

The best known evidence comes from Les Binet and Peter Field, who analysed more than 800 campaigns in the databank of the UK Institute of Practitioners in Advertising. A brand with a larger share of attention than of market tends to grow; the reverse tends to decline. As a rule of thumb 10 points of excess is worth roughly 0.5 percent of extra market share growth per year, and LinkedIn's B2B Institute puts business markets at seven tenths.

Now the limit, which almost every explainer leaves out. That research is about advertising spend, not editorial coverage. Carrying the conclusion over to media mentions is using evidence for something it is not about.

For attention that was not bought, separate evidence exists. You, Vadakkepatt and Joshi pooled 51 studies in Journal of Marketing and found that both the volume and the tone of online word of mouth relate to sales, with tone weighing more heavily. Being mentioned often in a negative context is not a gain.

Share of voice in SEO measures something else

In SEO, share of voice usually means your portion of the estimated clicks on a set of keywords, derived from positions and a click-through curve. Every rank tracker works that out its own way.

That is an estimate of traffic, not a count of attention. Put both figures in one report without naming the difference and you are comparing an estimate with a count. We do not calculate that variant.

Three figures that all go by share of voice
What it measuresWhere the number comes fromDo we calculate it?
In AI answersYour portion of the brands that get namedA count per answer, with every named brand in the denominatorYes
In SEOYour portion of the estimated clicks on a set of keywordsPositions times a click-through curve, different in every toolNo
Share of marketYour portion of the revenue in the categoryMarket figures or your own revenue, outside the monitoringNo

About the figures

The window, the denominator and the way we establish prominence are written out in the measurement protocol, including why we do not turn them into one composite figure.

How we calculate

Sources

The research this page refers to, with its scope attached. A figure from a US consumer survey is not the same as a figure about a business market, and that belongs on the page.

  1. 01
  2. 02
  3. 03

What Reptor offers here

Questions about the offering on this subject: what you get, what it costs, and where we stop.

Which competitors is it measured against?

Against every brand named in the answers and results we measure, not only a list you supply up front. You can also follow specific competitors, which gives them their own place in the dashboard with their own sentiment and visibility.

How many questions are run?

That depends on the plan. In practice five to seven questions containing your brand name are running, alongside the questions without a brand name on which share is measured, each across eight AI platforms.

Can I see which of my pages are cited?

Yes. That makes visible which of your own publications are actually used and which land nowhere.

Is this in both plans?

Share and prominence are in both. The number of search terms and tracked entities differs; the pricing page sets out what each plan includes.

What if my brand name means something else as well?

That is handled by the relevance filter, using a knowledge profile that separates your organisation from the namesake. It is work done up front, because a share that counts namesakes is unusable.

Frequently asked questions

Why show share and prominence separately?

Because being mentioned in passing everywhere is not the same as being named first once. Merging them into one number requires a weight we would have to invent, and then we would be publishing an assumption as a measurement.