Free AI Share of Voice Calculator

What is AI share of voice, and what does the number mean?

Share of voice is a ratio: how much of a category's attention goes to you. In AI answers the attention is a brand named in the reply. The arithmetic is simple. The hard parts are what you count, how many answers you count, and what you are allowed to say about the result.

From ad budgets to AI answers

Share of voice started as an advertising measure: a brand's share of all the ad spend or impressions in its category. Public relations teams borrowed it for press coverage, and social teams for mentions online. Each time the question was the same one. When people in this market hear about a brand, how often is it us?

AI assistants add a new place to ask it. Someone types "best fee-only financial planner in Denver" into ChatGPT or Perplexity, and the answer names three or four firms. Being one of them is the new version of being on the first page. AI share of voice is your share of the brand mentions across a set of those answers.

The share of voice formula

Every share of voice calculation is the same division. Count the mentions of your brand, count the mentions of every brand in the same answers including yours, divide the first by the second, and multiply by 100.

// share of voice formula
share of voice = your mentions ÷ all brand mentions × 100

// the worked example in the card above: 4 answers
Your firm   3     Rival A   4
Rival B     2     Rival C   1
total      10

share of voice = 3 ÷ 10 × 100 = 30%

That is all a share of voice calculator, or SOV calculator, does. This one adds two things the division leaves out: the other brands' share of the same total, and how far the result moves when one answer goes the other way, which is the part most people skip.

Start with a fixed prompt set

The number means nothing unless the questions stay the same. Change the prompts and the share moves because the questions changed, not because your visibility did. So the first job is a written list: the questions a buyer in your market would actually type, phrased the way they would phrase them, and saved so every later run asks the same thing.

Two rules keep the list honest. Leave your brand name out of the prompts, because a question that names you will mention you almost every time and inflate the share. Verand's dashboard offers an unbranded view for exactly that reason. And cover the whole buying journey, not only the questions you already win: "how do I choose", "what does it cost", "who is best for" and the comparisons.

What counts as a mention

A mention is an answer naming a brand in its text, counted once per answer. A reply that names Rival A three times still gave Rival A one slot in that answer, and counting all three would reward repetition rather than presence. Count common variants of each name the same way, such as an abbreviation or the name without "LLC", and decide before you start which brands are in the total.

That last choice changes the denominator. "All brand mentions" can mean every brand any answer named, or only you and the competitors you track. Verand's dashboard uses the second: your mentions against you plus your tracked competitors. Either works if you keep it fixed. Switching between them makes a trend line that measures your spreadsheet, not the market.

Share of voice is not visibility

Two numbers get confused here. Visibility, sometimes called mention rate, is the share of answers that name you at all. Share of voice is your share of all the brand mentions in those answers. The worked example shows how far apart they can sit.

A firm can be named in most answers and still hold a small share, because every answer names four rivals beside it. The reverse happens too, in a thin category where answers name one or two firms. Report both, and do not let a high visibility figure stand in for share.

Mentions and citations are different numbers

A mention is your name in the answer's words. A citation is your website among the sources the answer links to. They measure different things, and an assistant can do one without the other: it can name your firm and cite a directory, or cite your article and recommend a competitor. Verand records them separately for the same reason, and so should a spreadsheet. This calculator works on mentions. A citation share is the same division over cited sources instead of named brands.

Why the share differs by assistant

Run the same prompts through different assistants and the share will not match. Perplexity searches the web for most questions and shows its sources. Google AI Overviews and Google AI Mode are built on Google Search. ChatGPT, Gemini and Claude can answer from what the model already learned or search the web, depending on the question and the settings. Different sources produce different shortlists.

So keep a share for each assistant as well as the combined figure. A combined 25% can hide a strong position in one assistant and near-absence in another, and the fix for each is different. Pooling answers from several assistants into one total is fine if you keep the mix the same from one run to the next.

How many answers make a number

There is no official sample size, and any page that gives you one is guessing. What the arithmetic does tell you is how much each mention weighs. With 20 mentions in the total, one mention is 5 points of share. With 200, it is half a point. A move from 30% to 35% on a total of 20 is one answer going your way, which is well within the difference between two runs of the same prompt.

That run-to-run difference is real. The same prompt asked twice can produce two different shortlists, which is why careful writers on the subject call share of voice directional rather than exact. Treat a small sample as a rough reading, grow it with more prompts and more runs before you act on a small change, and use the calculator's "one more, one fewer" line to see whether a change is bigger than a single answer.

Tracking share of voice over time

The value is in the trend, not in one reading. For the trend to mean anything, four things have to stay the same between readings: the prompt list, the assistants, the set of brands in the total, and the counting rule. Compare equal windows, such as this month against last month, rather than a single day against a month, and note anything that changed on your side, such as a new page or a site change, so you can line it up against the curve later.

Before you publish the number

Most pages on this subject treat the share as a scoreboard to show off. For a regulated firm it is a factual claim, and factual claims in advertising need support. The FTC's position is that an objective advertising claim needs a reasonable basis before it is made. For registered investment advisers, the SEC's Marketing Rule (Rule 206(4)-1) bars an advertisement that includes a material statement of fact the adviser does not reasonably believe it can substantiate if the SEC asks. "We lead AI answers in our category", resting on a few dozen prompts you wrote yourself, is hard to support that way.

The safer use is internal: a number the team tracks to decide what to write next. If you do publish it, say what it rests on: the prompts, the assistants, the dates, the brands in the total and the counting rule. Then run the copy past whoever reviews your advertising. This page describes the question; it is not legal advice.

Common mistakes

• Your brand in the prompts. "Is Smith Wealth a good planner?" will mention Smith Wealth every time. Branded prompts belong in a separate view.
• Counting repeats. Three mentions of one brand in one answer is one mention. Counting all three turns the share into a measure of how wordy the answer was.
• A total without you in it. The denominator includes your own mentions. Leave them out and the share can pass 100%.
• A moving brand list. Adding a competitor to the total halfway through lowers everyone's share on the day you add it.
• Rerunning until it looks better. An unflattering answer is part of the sample. Rerunning it until you appear is how a 20% becomes a 40% on paper.
• Reading a small sample to the decimal. On a total of 20, the first decimal place is noise. Look at what one mention is worth before reading a change.
• Mixing mentions and citations. They are two numbers. A page cited in an answer that recommends a rival is not a win in the share of voice column.

What a useful share of voice figure looks like

A written prompt list with no brand names in it. The same assistants every time, with a share for each. A fixed set of brands in the total, and one mention per brand per answer. A total large enough that one answer does not swing the result by more than you care about, and readings compared in equal windows. None of that tells you why the share is what it is, and no calculator can. It tells you whether the share is moving, which is the question worth tracking.