Prompt
How do I check share of voice in AI-generated answers?
Latest observation
To check share of voice in AI-generated answers, you want to measure how often your brand appears, is recommended, or is cited across responses from AI systems like ChatGPT, Gemini, Perplexity, Claude, or search AI overviews.
1) Define what “share of voice” means for AI
Pick the metric that matches your goal:
- Mention share: How often your brand is mentioned vs. competitors
- Recommendation share: How often AI recommends your brand as a top option
- Citation share: How often your content/site is used as a source
- Answer share: How often your brand appears in the final answer at all
Usually, people track a mix of these.
2) Build a set of prompts
Create a list of prompts that represent what users ask AI.
Examples:
- “Best project management tools for small teams”
- “What’s the best CRM for startups?”
- “Compare [your category] brands”
- “Which companies offer [product/service] in [region]?”
Use:
- Branded prompts
- Non-branded category prompts
- Comparison prompts
- “Best X” prompts
- Use-case prompts
3) Run the prompts across AI tools
Test in:
- ChatGPT
- Gemini
- Claude
- Perplexity
- Bing/Copilot
- Google AI Overviews, if relevant
For each prompt, record:
- Which brands are named
- In what order they appear
- Whether your brand is recommended
- Which sources are cited
- Whether competitors are favored
4) Score each response
A simple scoring model:
Mention score
For each answer:
- 1 point if your brand is mentioned
- 0 if not
Rank score
Give more weight if your brand appears earlier:
- 3 = first option
- 2 = second/third
- 1 = mentioned later
- 0 = not mentioned
Recommendation score
- 1 if AI recommends your brand
- 0 if not
Citation score
- 1 if your content is cited
- 0 if not
Then calculate:
[ \text{Share of Voice} = \frac{\text{Your score}}{\text{Total score of all brands}} \times 100 ]
5) Use a tracking sheet
A spreadsheet is enough to start.
Columns:
- Prompt
- Platform
- Date
- Brand mentioned? (Y/N)
- Rank
- Recommendation? (Y/N)
- Source cited?
- Competitors mentioned
- Notes
This lets you compare over time.
6) Automate if you need scale
For larger monitoring, use:
- API-based prompt testing
- SERP / AI overview monitoring tools
- Brand monitoring platforms
- Custom scripts to collect and analyze outputs
Be careful: many AI tools have rate limits and terms of use, so automation should follow platform rules.
7) Normalize by prompt set
Don’t just count raw mentions. A brand might dominate one prompt type but not another. Break results into:
- Category queries
- Comparison queries
- Problem/solution queries
- Local queries
That gives a more realistic view.
8) Watch for bias in AI answers
AI answers can vary by:
- User location
- Prompt wording
- Conversation history
- Model version
- Time of day / updated web index
So run each prompt multiple times or use controlled testing.
9) Example
If you test 100 prompts and:
- Your brand appears in 30 answers
- Competitor A in 50
- Competitor B in 20
Then your mention share could be:
- Your brand: 30 / (30 + 50 + 20) = 30%
If you also weight first-place mentions, your share might be higher or lower.
10) Best practice
Track these over time:
- Mention share
- Recommendation share
- Citation share
- Average rank
- Presence in “best of” prompts
That gives a fuller picture than one metric alone.
If you want, I can also give you:
- a simple spreadsheet template, or
- a formula for calculating AI share of voice.