AI Answer Share of Voice: How Often Your Brand Appears in Relevant AI Answers

In traditional SEO you can say "we rank third for this keyword". AI answers don't have a stable ranking. There's no page one, only an answer that might include you, or might not, depending on the day. Share of voice is a way to bring order to that. It asks: out of all the relevant AI answers, in what proportion do we appear?

The basic formula

Share of voice = number of answers that include your brand / total number of answers tested

Run it for each competitor too, and you get comparable numbers.

Brand	Answers including brand	Total answers	Share of voice
You	18	120	15%
Competitor A	54	120	45%
Competitor B	30	120	25%
Competitor C	12	120	10%

Percentages don't need to add up to 100, because one answer can mention several brands.

Step 1: define the space

Decide which topic you're measuring. "Website audits for small businesses" is a space. "Marketing" is too broad to be useful. Choose a handful of competitors who genuinely overlap with you.

Step 2: build a controlled prompt set

Write prompts that cover the full buying journey:

• Early research: "how do I check if my website has SEO problems"
• Consideration: "what should I look for in a website audit service"
• Comparison: "website audit tool vs hiring a consultant"
• Decision: "best website audit service for a small agency"

Use the same list each time you measure. Changing the prompts changes the result, so version your set and note any edits.

Step 3: control the conditions

• Test on the same platforms each time
• Use fresh sessions, logged out where possible
• Run each prompt at least three times
• Record the date and the platform

Step 4: record the data

For each run, note which brands appeared, in what position (first named, or later), and whether your domain was cited. Position matters, since the first recommendation tends to attract more attention.

You can extend the basic measure with:

• Weighted share, giving more credit to first mentions
• Citation share, counting only answers that link to the brand's site, see mention vs citation
• Sentiment, flagging mentions that are negative or caveated

Step 5: report by segment

Overall numbers hide detail. Split by prompt type, by platform and by topic. You may find that you dominate early research but disappear at the decision stage, which suggests a different problem to fix.

Step 6: use it to decide

A share-of-voice report is useful only if it changes what you do. Link each weak segment to a possible cause:

• Low on comparison prompts: you may lack comparison content, see AI comparison readiness
• Low on recommendation prompts: look at recommendation readiness
• Low everywhere: check crawler access and entity clarity first

Honest limits

• Different platforms use different data and sources, so a number from one doesn't transfer to another.
• Personalisation and location can shift results.
• Platforms change quickly, so a drop might not be your fault.
• A sample of 100 answers has real uncertainty. Don't report one decimal place.

Common mistakes

• Picking prompts that favour your brand, which inflates the number.
• Switching tools between months and comparing the results.
• Only measuring branded prompts, where you'll always win.
• Ignoring competitors, so the number has no context.

Measured carefully, share of voice gives a stable reference point in an unstable channel.

Further reading: Google's page on AI features and your website and the Search Console performance report are good places to check the details straight from the source.