AI Share of Voice: Definition, Formula and Reporting Template
Calculate mention, citation and recommendation share of voice using transparent formulas, competitor definitions and a repeatable reporting template.
AI share of voice measures a brand’s relative presence within a defined set of AI-assisted search observations. Instead of asking only “Were we mentioned?”, it asks “How much of the visible brand or citation space did we receive compared with the alternatives?”
The metric is useful for competitor context, but it is easy to misuse. AI answers vary, platforms expose different data and one prompt can name several brands or sources. A defensible report must define the universe, outcome and denominator precisely.
The Short Answer
Use three separate share-of-voice metrics:
- Mention Share of Voice: Your brand mentions divided by all tracked brand mentions.
- Citation Share of Voice: Your visible citations divided by all tracked visible citations.
- Recommendation Share of Voice: Your appropriate recommendations divided by all tracked appropriate brand recommendations.
For example:
Mention SOV = Your mentions ÷ Mentions of all tracked brands × 100
Do not combine the three without showing the components. A brand may lead in citations while a competitor leads in recommendations.
Why AI Share of Voice Is Different From Traditional SOV
Traditional media share of voice often compares brand mentions, advertising spend or impressions within a channel. AI answers introduce new complications:
- One response can mention several brands.
- A source can be cited without the brand appearing in the prose.
- A brand can be recommended without an owned citation.
- The same prompt can produce different results across runs.
- Platforms, locations and conversation states are not directly equivalent.
- A mention can be negative, inaccurate or irrelevant.
The metric should therefore be described as share within a tracked observation set—not total market visibility.
Step 1: Define the Competitive Universe
Create the competitor set before measuring.
Include:
- Direct commercial competitors.
- Alternatives buyers genuinely consider.
- Platforms or marketplaces that compete for the same attention.
- Unprompted brands that repeatedly appear during a discovery phase.
Avoid adding dozens of marginal brands. A stable set of five to ten competitors is manageable for many programs.
Document additions and removals. Changing the competitor set changes the denominator, so the new period may not be comparable with the old one.
Use the AI brand visibility guide to define the brand and competitor evidence consistently.
Step 2: Build the Prompt Universe
AI share of voice depends as much on the prompt mix as the competitor set.
Group prompts by:
- Audience.
- Country and language.
- Topic.
- Funnel stage.
- Commercial intent.
- Product or service line.
A provider could dominate broad educational prompts and disappear from vendor-selection questions. Always report SOV by intent before presenting an overall number.
Keep branded prompts separate from non-branded discovery prompts. If you ask about your brand directly, the resulting mention should not be treated as competitive discovery.
The AI visibility tracking system provides the observation protocol.
Step 3: Record the Right Outcomes
For each response, record every tracked brand and source.
Mention event
Count a mention when the answer explicitly names the brand or a defined product name. Decide how alternate names and subsidiaries are handled.
Citation event
Count a citation when a visible source link belongs to the tracked domain. Define subdomains and regional domains in advance.
Recommendation event
Count a recommendation only when the answer suggests the brand as suitable for the prompt’s audience or need. A brand listed as an unsuitable example is not a positive recommendation.
Prominence
Optionally record position, section or answer prominence. Keep raw position separate from the main count unless the weighting rule is explicit.
The Three Core Formulas
Mention Share of Voice
Your brand mentions ÷ Total tracked-brand mentions × 100
Suppose five brands receive 200 mentions across the observation set and your brand receives 44.
44 ÷ 200 × 100 = 22% Mention SOV
Citation Share of Voice
Your domain citations ÷ Total tracked-domain citations × 100
If tracked domains receive 160 citations and your domain receives 32:
32 ÷ 160 × 100 = 20% Citation SOV
Recommendation Share of Voice
Your appropriate recommendations ÷ Total appropriate tracked-brand recommendations × 100
If the set contains 80 recommendation events and your brand receives 12:
12 ÷ 80 × 100 = 15% Recommendation SOV
The pattern matters: 22% mention share, 20% citation share and 15% recommendation share suggest that the brand is visible but less competitive at the recommendation stage.
Alternative: Prompt-Level Share
Event-based SOV can be dominated by long answers that list many brands. Prompt-level share answers a different question:
Prompts where your brand appeared ÷ Prompts where any tracked brand appeared × 100
Use it to measure coverage across the prompt set. Report it as prompt presence share, not citation share.
Should Position Be Weighted?
A weighted model can assign more points to earlier or more prominent appearances. For example:
- First named brand: 3 points.
- Second: 2 points.
- Later appearance: 1 point.
This can be useful, but the order of brands in a generated answer is not necessarily a stable ranking. Research shows AI answer visibility varies across samples. Report the unweighted SOV beside any weighted result and avoid pretending the weights come from the platform.
Use Priority Weights
Prompt importance may deserve weighting. Assign points such as critical, important and monitoring, then calculate achieved share from the available weighted events.
Do not let a few prompts overwhelm the report. Show:
- Weighted overall SOV.
- Unweighted overall SOV.
- SOV by intent.
- Observation count.
Bing Citation Share vs. Your Tracked SOV
Bing Webmaster Tools has introduced Citation Share within its AI Performance reporting. Bing defines it as the percentage of citations attributed to your site out of all citations for a specific grounding query in supported experiences.
That first-party metric is not interchangeable with a cross-platform score you calculate from prompts. Keep the labels distinct:
- Bing Citation Share: Bing’s aggregated metric for a grounding query.
- Tracked Citation SOV: Your calculated share across the defined observation set.
Bing also notes that Citation Share is not a ranking, traffic measure or quality score. Changes can reflect demand, freshness, model behavior and changes across the web.
Account for Uncertainty
AI share of voice is a sample estimate. Identical prompts can produce different brands and sources.
Improve reliability by:
- Repeating important prompts.
- Maintaining stable conditions.
- Reporting observation counts.
- Calculating confidence intervals for larger samples.
- Showing multi-period trends instead of isolated points.
- Marking changes in prompt mix or competitor set.
Research on AI visibility uncertainty found that apparent differences between domains may fall within the normal noise of repeated sampling. Do not declare a competitive win from a one-point change without enough observations.
Reporting Template
Executive summary
- Overall mention, citation and recommendation SOV.
- Change versus previous period.
- Strongest and weakest intent groups.
- Material accuracy or reputation issues.
- Top actions.
Competitive table
| Brand | Mention SOV | Citation SOV | Recommendation SOV | Accuracy rate |
|---|---|---|---|---|
| Your brand | 22% | 20% | 15% | 90% |
| Competitor A | 26% | 18% | 25% | 88% |
| Competitor B | 18% | 24% | 20% | 92% |
| Other tracked brands | 34% | 38% | 40% | — |
Intent breakdown
Report educational, problem-solving, vendor-discovery, comparison and validation prompts separately.
Source breakdown
List owned pages and third-party sources most frequently cited with each brand. This turns the comparison into an evidence plan.
Action queue
Every finding should have an owner, due date and retest date.
How to Interpret Common Patterns
High mention share, low citation share
The brand is named but owned evidence is not being visibly sourced. Review original data, methodology pages, source clarity and technical eligibility.
High citation share, low recommendation share
The domain is useful for information but not competitive in buyer-selection answers. Improve commercial-fit content, proof and third-party corroboration.
Low mention share, high accuracy
The brand is described well when present but lacks reach across the prompt set. Strengthen topic association and relevant authority.
Rising SOV, falling referral traffic
The visible answer mix, query demand or landing-page experience may have changed. Do not assume SOV and traffic move together.
Use the AI visibility optimization framework to assign the correct workstream.
AI Share of Voice vs. AI Visibility Score
An AI visibility score evaluates your brand across presence, citation, recommendation and accuracy. AI share of voice is comparative: it measures your portion of tracked brand or citation events.
Use both when needed:
- Visibility score: “How complete and accurate is our performance?”
- Share of voice: “How does our presence compare with the defined alternatives?”
Do not merge them into one number unless the combined formula is genuinely useful and fully documented.
Common Mistakes
- Calling a single answer market share.
- Mixing mentions, citations and recommendations.
- Changing competitors without restating the baseline.
- Including branded prompts in competitive discovery SOV.
- Counting negative mentions as positive recommendations.
- Comparing unlike platforms and markets.
- Hiding sample size.
- Overweighting answer order.
- Treating Bing Citation Share as a universal cross-platform metric.
- Reporting small changes without uncertainty.
Final Takeaway
AI share of voice is valuable when the scope is explicit. Define the competitors and prompt universe, count mentions, citations and recommendations separately, repeat the observations and report the method beside the result.
The objective is not to manufacture a large percentage. It is to understand where competitors earn visibility, which sources support them and what evidence or experience your audience still needs.
For a structured competitive baseline, use the AI Visibility Audit.
Frequently Asked Questions
What Is AI Share of Voice?
It is a brand’s percentage of tracked mentions, citations or recommendations within a defined competitor and prompt set.
Is AI Share of Voice the Same as Market Share?
No. It measures observed answer visibility, not revenue, customers or total demand.
How Often Should SOV Be Measured?
Monthly is suitable for many programs. Use a stable prompt set and repeated observations; report weekly only when the business needs a faster campaign or risk-monitoring cadence.
Can SOV Exceed 100%?
No, when calculated correctly within one event type. All brand shares for that denominator should total 100%, allowing for rounding.
