Prompt
How do I measure share of voice in AI-generated answers?
Latest observation
To measure share of voice (SoV) in AI-generated answers, think of it as:
How often your brand is mentioned, recommended, or cited compared with competitors when users ask relevant questions in AI tools.
Because AI answers are probabilistic and can vary by prompt, model, and source context, SoV in AI is usually measured with a tracked prompt set rather than one single query.
1) Define what “share of voice” means for AI
You need to choose the unit you’ll measure:
- Mention share: how often your brand appears in an answer
- Recommendation share: how often your brand is recommended as a top option
- Citation share: how often your site/content is cited as a source
- Answer-position share: whether you appear first, in a shortlist, or not at all
Most teams track all four.
2) Build a representative prompt set
Create a list of prompts that reflect real user intent, such as:
- “Best CRM for small business”
- “Compare [category] tools”
- “What is the best [product type] for [use case]?”
- “Alternatives to [competitor]”
- “Which companies offer [feature]?”
Group prompts by:
- category
- funnel stage
- geography
- language
- brand vs non-brand intent
3) Run prompts across AI surfaces
Test in the places that matter to you:
- ChatGPT
- Perplexity
- Gemini
- Claude
- Copilot
- Google AI Overviews / AI Mode if relevant
- vertical assistants in your industry
Use a consistent methodology:
- same prompt wording
- same locale/language
- same model/version if possible
- multiple runs per prompt to account for variability
4) Score the answers
For each response, record:
- Was your brand mentioned? Yes/No
- Was it recommended? Yes/No
- Was it cited? Yes/No
- What rank/position did it have?
- Was the mention positive/neutral/negative?
- Which competitors appeared?
Then calculate metrics like:
Mention SoV
[ \text{Mention SoV} = \frac{\text{# prompts where your brand is mentioned}}{\text{# total prompts}} \times 100 ]
Recommendation SoV
[ \text{Recommendation SoV} = \frac{\text{# prompts where your brand is recommended}}{\text{# total prompts}} \times 100 ]
Competitive share among mentions
[ \text{Brand Share} = \frac{\text{Your brand mentions}}{\text{All brand mentions across the set}} \times 100 ]
Example:
- Your brand appears in 42 answers
- Competitor A appears in 60
- Competitor B appears in 18
- Total brand mentions = 120
Your brand’s competitive SoV: [ 42 / 120 = 35% ]
5) Weight by importance
Not all prompts are equal. You may want to weight by:
- search volume
- conversion intent
- revenue potential
- market segment importance
For example, a high-intent prompt like “best enterprise analytics platform” may matter more than a top-of-funnel “what is analytics?”
6) Track citations separately from mentions
In AI search, being cited can be as important as being named. Track:
- citation rate
- citation rank/order
- whether your domain is used as evidence
- whether citations come from your own pages or third-party sources
A brand can have low mention SoV but high citation SoV if its content is frequently used as a source.
7) Measure over time
Because AI answers change, SoV should be tracked as a trend:
- weekly or monthly
- by model
- by prompt cluster
- before/after content or PR changes
This helps you see whether:
- new content improved visibility
- competitors gained ground
- a model update changed answer patterns
8) Use a simple dashboard
A useful AI SoV dashboard usually includes:
- total prompts tracked
- mention SoV
- recommendation SoV
- citation SoV
- average rank/position
- competitor comparison
- prompt cluster breakdown
- trend line over time
9) Be aware of limitations
AI SoV is not as deterministic as classic SEO because:
- answers vary by model and session
- source retrieval can change
- personalization and location can affect results
- some tools hallucinate or omit brands unpredictably
So it’s best to report SoV as:
- sample-based
- by platform
- by prompt set
- with confidence over time
Practical formula to start with
If you want a basic version, start with:
- Pick 50–100 important prompts
- Run them 3 times each in 2–3 AI tools
- Record whether your brand appears
- Calculate:
- mention rate
- recommendation rate
- citation rate
- competitor share
Example
Suppose you track 100 prompts in ChatGPT:
- Your brand is mentioned in 28
- recommended in 15
- cited in 10
- total competitor mentions across all answers = 80
- your brand mentions = 28
Then:
- Mention SoV = 28%
- Recommendation SoV = 15%
- Citation SoV = 10%
- Brand share of all mentions = 28 / (28 + 80) = 26%
If you want, I can also give you:
- a spreadsheet template for tracking AI share of voice, or
- a step-by-step measurement framework for ChatGPT/Perplexity/Gemini.