Share of Voice in AI Answers: How to Read It and What to Do Next

Share of voice is the metric most dashboards put first, and the one most teams misread. It is a ratio. It says how much of the conversation your brand owns compared with the brands around it. It does not say whether the conversation is good, whether it converts, or whether you can change it.

Used well, it is a steering wheel. Used badly, it is a number to celebrate on slow weeks.

What it measures

In AI search, share of voice compares your brand's presence across a set of tracked prompts against competitors' presence in the same answers. If a hundred responses name four brands in total and yours appears in thirty of the mentions, that is a share of thirty percent. The exact formula varies by tool, so check the definition before comparing numbers between products.

Two things make AI share of voice different from the classic version in search rankings. First, answers are generated, so the brand list can change between runs. Second, many answers name only two or three brands, which makes each slot more valuable than a position on page one.

Share of voice is not the same as a mention rate

A mention rate asks, in what percent of responses did we appear? Share of voice asks, among all the brand mentions, how many were ours? You can have a high mention rate and a modest share, if the answers name many brands. You can also have a small mention rate and a large share, if the topic is niche and few brands are named at all.

Keep both on the dashboard. The gap between them tells you about crowding.

Where the number goes wrong

The prompt set decides everything. Share of voice is only as good as the prompts behind it. A set full of branded questions will show inflated numbers, because people who name you get you. Keep organic, brand, and competitor-comparison prompts labeled and report them separately.

Competitors must be real. If the competitor list leaves out the brand that actually shows up in answers, your share looks better than it is. Promptwatch auto-discovers competitors from the responses, which helps catch the ones you forgot.

Brand matching needs rules. What counts as a mention? A product name, a nickname, a misspelling? Promptwatch lets you set brand aliases and a Brand Book whose analysis instructions decide what counts as a mention, so the count follows your definition rather than a default.

Averages hide the split. A fifteen percent share across ten engines could be thirty percent in one and near zero in another. Look at the breakdown by model, country, and topic.

The views that make it useful

Promptwatch reports Visibility Score, Share of Voice, Sentiment, and Position in Answer together, and the heatmap compares any two dimensions, such as brands against models or topics against countries. That heatmap is the quickest way to find a pocket where you are weak. A competitor heatmap does the same for rivals, and A/B period comparison shows what changed between two date ranges.

A few readings to practice:

• Topic by model. Where is your share lowest? That is where the content gap is probably widest.
• You against the top competitor, by country. If you win in one country and lose in another, the cause is often local sources.
• Before and after a publish date. Did the share move for the topic you published on, and nothing else?

What decisions it should drive

Where to publish. Pick the topics with the lowest share and the highest search volume. Promptwatch shows prompt volumes and difficulty alongside citation rates, so you can weigh effort against upside.

What to refresh. If a topic had a good share and then dropped, check the pages that held it. Citation Trends reports per-page decay, so you can tell whether the drop is a slow fade or a sudden break.

What to fix on the technical side. If the share is low on an engine and the relevant bot is missing from your crawler logs, start there. No amount of writing helps a page that cannot be fetched.

What to tell leadership. Share of voice is a good headline. Pair it with a traffic number from visitor analytics so the story has an outcome attached.

A weekly reading routine

1. Open the dashboard for the last seven days and compare to the previous seven.
2. Look at the heatmap for the largest swing.
3. Open the prompts behind that swing.
4. Read three of the actual responses, not just the numbers.
5. Write one action and one hypothesis.

Step four is the one people skip. Reading a few real answers will tell you more than any chart about why the number moved. Maybe a new competitor article appeared, or a Reddit thread started getting cited.

Limits worth stating

Share of voice cannot tell you whether the mentions persuaded anyone. It cannot tell you whether a user clicked. It is a visibility measure, not a revenue measure. Pair it with visitor analytics for AI-referred traffic and conversions, which Promptwatch attributes through a lightweight script, the GTM template, or the WordPress plugin.

The recommendation

Treat share of voice as the question and the rest of the platform as the answer. Promptwatch lets you define what counts as a mention, break the number down by model and topic, and then move to the pages and crawler data behind it. If you only have a basic tracker today, ask whether it lets you click from the ratio to the underlying responses and pages. If it does not, the number is a thermometer without a doctor.