TL;DR
| Answer Share of voice = your metric ÷ the total for all tracked brands × 100. Use one metric per channel. |
Channels Paid, organic search, social, PR and AI answers each get their own numerator and denominator. |
| AI answers Count citations first, mentions second, per engine, on a fixed prompt set with repeated runs. |
Example Four brands, three engines, 1,040 citations. The brand in the example holds 29.4% pooled. |
| Good SOV There is no universal benchmark. Compare it with your market share and an even split of the field. |
Avoid Single runs, branded prompts, changing the competitor set and comparing numbers across tools. |
To calculate share of voice, divide your brand’s number for one metric by the total for all the brands you track, then multiply by 100. The metric changes by channel, such as impressions, clicks, mentions or citations, but the share of voice formula never does.
This guide gives the formula, the numerator and denominator for each channel, and a step-by-step method for share of voice in AI search. It ends with a worked example across three engines, a view on what a good number looks like and the mistakes that break the maths.
1How do you calculate share of voice?
Pick one metric, one set of competitors and one time window. Add up your number and every competitor’s number, then work out your slice of the total.
Formula: share of voice
SOV % = your metric ÷ total market metric × 100
The total market metric is your number plus every tracked competitor’s number, for the same channel, set and dates.
A quick share of voice calculation looks like this. Your brand earns 12,000 organic clicks on a tracked keyword set. All tracked brands together earn 60,000. Your share of voice is 12,000 ÷ 60,000 × 100, which is 20%.
Share of voice is not the same as share of search. Share of search measures how much people look for your brand, while share of voice measures how visible you are across channels. Our share of search guide covers the first, and the two work best together.
2How do you calculate share of voice by channel?
Each channel has its own metric, so the denominator changes while the formula stays the same. Keep the competitor set identical across channels, or the numbers cannot be compared.
| Channel | Your metric | Total market metric | Typical data source |
|---|---|---|---|
| Paid search | Your ad impressions | Impressions you could have received | Auction insights in your ad platform |
| Organic search | Your impressions or clicks on a tracked keyword set | Impressions or clicks of all tracked brands on that set | Rank and visibility tracking |
| Social | Mentions of your brand, or engagements on those posts | Mentions or engagements for all tracked brands | Social listening |
| PR and earned media | Media mentions of your brand | Media mentions for all tracked brands | Media monitoring |
| AI answers | Citations of your domain in AI answers | Citations of all tracked brands on the same prompts | A prompt-based citation tracker |
For paid search, the denominator is the impressions you were eligible for, not all impressions in the market. For social, count untagged text mentions as well as tags, and clean out spam, bots and duplicates before you divide.
3How do I calculate share of voice in AI answers?
Share of voice in AI search works like every other channel, with one big difference. The answer changes every time you ask, so a single run proves nothing. You need a fixed prompt set, repeated runs and one engine at a time.
Definition: AI share of voice
Your slice of the citations among all tracked brands, on the same prompts, on the same engine. A citation means the answer links to or credits your domain. A mention means the answer names your brand, with or without a link.
Lead with citations. A citation is what can send a visitor to your site, while a name drop sends nobody. Track mentions as a second number, because the gap between the two shows where AI engines know you but do not credit you. Our guide to brand performance metrics in AI search defines both rates.
| 1 | Build a fixed set of 30 to 50 prompts that a buyer would ask, with no brand names in them. |
| 2 | Choose three to five competitors and keep the same set every time you measure. |
| 3 | Pick the engines that matter to you, such as ChatGPT, Google AI Overviews, Google AI Mode, Perplexity and Gemini. |
| 4 | Run every prompt several times on each engine, in the same window and the same country. |
| 5 | Record citations and mentions for each brand in every answer. |
| 6 | Pool the counts for each engine, divide your count by the total and multiply by 100. |
| 7 | Repeat on a schedule and judge the trend, not one snapshot. |
Formula: AI share of voice
AI SOV % = your citations ÷ citations of all tracked brands × 100, per engine, on the same prompt set.
Swap in mentions for the same calculation on named-but-not-credited visibility.
How many runs is enough? A 2026 study from the University of St. Gallen found that the sources an engine cited overlapped by only 34 to 42% from one day to the next, and recommended at least seven runs per prompt per day (Schulte, Bleeker and Kaufmann). Treat that as a floor for any AI share of voice marketing metrics you report.
4What does a worked AI share of voice example look like?
Here is the method on one brand and three competitors. The numbers are illustrative and anonymised, built to show the arithmetic, not taken from a real account. The setup is 40 prompts, five runs per prompt on each engine, which gives 200 answers per engine.
Citations by engine
| Brand | ChatGPT | Perplexity | Google AI Overviews | Pooled total |
|---|---|---|---|---|
| Your brand | 96 | 150 | 60 | 306 |
| Competitor B | 140 | 120 | 100 | 360 |
| Competitor C | 84 | 90 | 70 | 244 |
| Competitor D | 60 | 40 | 30 | 130 |
| All tracked brands | 380 | 400 | 260 | 1,040 |
Share of voice by engine
| Brand | ChatGPT | Perplexity | Google AI Overviews | Pooled |
|---|---|---|---|---|
| Your brand | 25.3% | 37.5% | 23.1% | 29.4% |
| Competitor B | 36.8% | 30.0% | 38.5% | 34.6% |
| Competitor C | 22.1% | 22.5% | 26.9% | 23.5% |
| Competitor D | 15.8% | 10.0% | 11.5% | 12.5% |
On ChatGPT, 96 ÷ 380 × 100 gives 25.3%. On Perplexity, 150 ÷ 400 gives 37.5%. The pooled figure is 306 ÷ 1,040 × 100, which is 29.4%. Pool the counts, then divide. Averaging the three percentages would give 28.6% and misstate the result, because the engines have different totals.
The per-engine view is where the work starts. The brand leads on Perplexity but trails Competitor B on ChatGPT and Google AI Overviews. Where one competitor leads, look at the pages the engine cites for them and ask what claim those pages own.
Citations versus mentions on one engine
| Brand (ChatGPT) | Mentions | Mention SOV | Citations | Citation SOV |
|---|---|---|---|---|
| Your brand | 120 | 30.0% | 96 | 25.3% |
| Competitor B | 150 | 37.5% | 140 | 36.8% |
| Competitor C | 80 | 20.0% | 84 | 22.1% |
| Competitor D | 50 | 12.5% | 60 | 15.8% |
Mention SOV is 30.0%, but citation SOV is 25.3%. The engine names the brand more often than it credits it. That gap is the number to close, because mentions without citations send no one to your site.
5What is a good share of voice?
No benchmark applies to every market, and the reference guides we checked give none. A good share of voice is one that is high relative to your competitors, relative to your market share, and moving in the right direction over time.
Formula: excess share of voice
Excess SOV = share of voice − share of market
A planning rule of thumb, not a guarantee. Share of market means your share of category sales.
| Your position | What it can mean | What to do |
|---|---|---|
| SOV above market share | You are gaining attention ahead of sales, so growth may follow | Protect the lead and check that visibility turns into traffic |
| SOV close to market share | Visibility and sales are in step | Look for the claims where one rival leads |
| SOV below market share | Sales may stall as rivals take the attention | Raise investment in the channels where you trail |
For AI answers, add a simple second yardstick. With four tracked brands, an even split is 25% each. In the example, 29.4% pooled is above an even split, but the brand is below it on ChatGPT and Google AI Overviews. Judge each engine on its own before you report one blended figure.
6What mistakes break a share of voice calculation?
| Mistake | Why it distorts the number | Fix |
|---|---|---|
| Single runs | AI answers change every time. One run can swing your share up or down by chance. | Run each prompt several times per engine and pool the counts |
| Mixing in branded prompts | A prompt that names your brand makes you appear by design and inflates your share. | Keep the prompt set unbranded and fixed |
| Cross-tool comparisons | Tools sample prompts, engines and countries differently, so their numbers do not match. | Compare only numbers from the same tool, set and settings |
| A changing competitor set | Adding or removing a rival changes the denominator and breaks the trend. | Freeze the set, and restart the trend if you change it |
| Counting mentions as wins | A mention can be negative or carry no link. A spike can be backlash. | Lead with citations and review sentiment |
| Unclean data | Spam, bots and duplicates inflate counts, mostly on social. | Filter before you divide |
| Reading one snapshot | Share of voice is a leading signal, not proof of business impact. | Track the trend over several months and tie it to traffic |
7How can I cross-check my competitor set for free?
Our free AI competitor analysis compares how an AI engine describes your brand with how it describes up to three rivals. You enter your domain and the rivals, choose one engine and one country, and it ranks the claims where you trail, with the brand that leads each one. No account is needed for the free verdict, and a work email opens the full report.
Each run covers one engine, so repeat it for ChatGPT, Gemini and Perplexity. It is a snapshot of what each engine says about the brands, not a share of voice number, so use it to find the claim a rival owns, then check that claim against your citation counts.
What only Wellows adds
Running this by hand is slow, and the answers keep moving. The Wellows AI Visibility Score does the pooling for you. It shows the percentage of total citations your brand earns against your direct competitors, for the prompts you track, combined or split by ChatGPT, Google AI Overviews, Google AI Mode, Perplexity and Gemini. A leaderboard ranks you against your rivals and shows whether you are gaining or losing ground, and the dashboard refreshes every 24 hours.
It also counts implicit citations, which are the times you are not cited but another page about your topic is. This is your shadow reputation, and it is the narrative that AI engines create based on pages you have not written. A hand-built count only sees the pages that name you. Wellows shows the whole citation pool, so you can see who holds the narrative and by how much.
You can drill into the score by prompt, platform, competitor and region, which is the per-engine view from the worked example, kept current. Track it, fix the gap, then measure again.
A share of voice checklist for this week
- Choose one channel and the metric you will count.
- Freeze a competitor set of three to five brands.
- Write 30 to 50 unbranded prompts for AI answers.
- Run each prompt several times on each engine you care about.
- Pool the counts, then divide and multiply by 100.
- Compare the result with your market share and with an even split.
- Repeat next month with the same set, prompts and settings.
Conclusion
Share of voice is your metric divided by the total for the brands you track, times 100. In AI answers, count citations before mentions, pool repeated runs per engine and hold the prompt set and competitor set still. Start with one engine this week, then add the rest and watch the trend.
