Share of Voice in PR: Definition, formula, and examples

In this article

Share of Voice, often shortened to SOV or SoV, is the percentage of a defined coverage universe attributed to one brand, organization, product, or issue. It compares relative presence. It does not measure audience opinion, market share, or PR impact.

The phrase "defined coverage universe" carries most of the work. A Share of Voice value is only meaningful when the report names the competitors, topic, channels, market, period, counting unit, and any weighting.

The basic Share of Voice formula

For a mention-based analysis:

Share of Voice = eligible brand mentions ÷ eligible mentions of all compared brands × 100

Suppose a six-week media analysis finds 42 mentions of your brand and 58 mentions of the selected competitors. The comparison set contains 100 eligible brand mentions.

Your Share of Voice is 42 ÷ 100 × 100 = 42%.

This is an output metric. The AMEC Taxonomy of Evaluation places Share of Voice among measures of communication output. The figure shows how coverage was distributed. It does not show what an audience noticed or believed.

Decide what one unit means

SOV can use different units. Each produces a valid but different answer.

Do not switch units between reporting periods. A mention-based value cannot be compared directly with a reach-weighted value.

One article can count for several brands

Decide whether the analysis counts media items or brand appearances.

If one comparison article names three tracked companies, an item-based method may give one item to each company. The total number of brand-attributed items will then exceed the number of unique articles. That is not an error if the method is stated.

Another method can assign the article only to its dominant brand. That prevents multiple attribution but adds a subjective coding decision. Neither rule is universal. Write the rule before coding and keep it stable.

Mention-based and exposure-weighted SOV

Mention-based SOV gives every eligible unit the same weight. It is simple to reproduce and works well for volume comparisons.

Exposure-weighted SOV uses an audience estimate:

Exposure-weighted SOV = estimated exposure attributed to the brand ÷ estimated exposure for all compared brands × 100

This can distinguish a national publication from a small local outlet. It also inherits every weakness in the exposure data. Monthly unique visitors, print circulation, social impressions, and broadcast audience estimates are not interchangeable. Combining them requires a disclosed normalization method.

Never describe exposure-weighted SOV as verified attention. A potential audience can include duplication across outlets and people who never encountered the coverage.

Quality-weighted Share of Voice

A team may want target media, strong prominence, favorable tone, or key-message presence to count more heavily. That creates a custom index, not standard SOV.

A transparent version gives each item a published weight and then uses the weighted sum in the numerator and denominator. For example, target media might receive a weight of 2 and other eligible media a weight of 1.

Quality weighting can answer a useful management question, but it becomes easy to manipulate. Report the raw SOV beside it and publish every factor and weight.

Share of Voice and sentiment

Raw SOV counts presence regardless of tone. A crisis can increase a company's Share of Voice sharply.

That does not make SOV useless. It means the report needs a second dimension. Show positive, neutral, and negative coverage within each brand's share, or place the SOV trend beside a separate sentiment analysis.

Avoid multiplying SOV by sentiment and calling the result reputation. The combination is still a content-analysis index. Audience attitude requires audience research.

Share of Voice in AI answers

AI Share of Voice applies the same idea to a controlled answer sample. Its denominator is different from media SOV.

In aclipp, a brand has a presence run when an eligible AI answer contains a detected text mention or cites a configured brand domain. A run counts once for that brand even if the answer mentions it several times.

AI Share of Voice = presence runs for the brand ÷ presence runs for all tracked brands × 100

One answer can contribute one appearance to several brands. If the tracked brands produce 80 total presence runs and your brand accounts for 20, your AI Share of Voice is 25%.

Some teams call this Share of Model. The alternate label does not change the method. The report must still show the prompt library, providers, competitors, countries, languages, dates, and treatment of failed or empty responses.

AI SOV should be read with visibility rate. Visibility rate uses all eligible answer runs as its denominator. SOV uses appearances of the tracked brands. A brand can increase its visibility rate while losing SOV if competitors improve faster.

For the full measurement system, see What is AI visibility?.

What belongs in every SOV report

Publish these details with the percentage:

1. the topic, query, or campaign universe;
2. every included brand or competitor;
3. market and language;
4. channels or AI providers;
5. start and end date;
6. the unit of analysis;
7. weights, exclusions, missing data, and duplicate rules.

Changing the competitor set changes the result even when coverage stays identical. Keep the set stable for trend reporting, or mark the point where it changed.

How to interpret the result

Share of Voice answers one narrow question: how much of the defined conversation belonged to the brand? It does not establish:

• awareness or recall;
• positive reputation;
• preference or purchase intent;
• market share;
• communication impact;
• return on investment.

Connect the result to tone, message content, audience research, web behavior, or business outcomes according to the original objective.

Our PR KPI guide places Share of Voice in the broader measurement chain. For message content, continue with key message pull-through.