Most teams guess at what AI engines want. A Share of Answer study replaces the guessing with a fixed, repeatable measurement: the same buyer questions, asked of the same engines, on a schedule, with every cited source logged.
This is the method we use at the start of every engagement. You can run a lighter version yourself in a spreadsheet.
1. Build a commercial prompt set
Start from revenue, not keywords. List the questions a buyer asks in the weeks before they shortlist vendors:
- Best-of prompts: "best CRM for a 20-person SaaS sales team"
- Alternatives: "HubSpot alternatives for startups"
- Comparisons: "Loopwise vs Northpipe"
- Fit questions: "which payroll tool works for teams in India and the UAE?"
- Trust questions: "is [brand] worth the price?"
Thirty prompts is enough to see a pattern. A hundred gives you a baseline you can report on. Write them the way people actually type, including the messy ones.
2. Fix the engines and the conditions
Pick the engines your buyers use, usually ChatGPT, Gemini, Perplexity and Google AI Overviews. Keep the conditions stable from run to run: logged out where possible, the same location, no conversation history. AI answers vary between runs, so consistency in how you ask is what makes month-over-month change meaningful.
3. Log four things per answer
- Appears: is your brand named at all?
- Recommended: is it presented as a good choice, not just mentioned?
- Cited: does the engine link to one of your pages?
- Sources: which URLs does the engine cite, yours or anyone else's?
The sources column is the most valuable one. It tells you which comparison sites, directories, forums and publications the engines lean on in your category.
4. Turn the log into a gap map
Group the prompts you lose by why you lose them. Is a competitor cited from a listicle you're not on? Is there no page on your site that answers the question directly? Do the engines describe your product inaccurately? Each answer points to a different fix: off-site citation work, a new Citation Asset, or entity and schema work.
5. Re-run on the same set
Run the identical prompt set again every month. Your Share of Answer is the percentage of those prompts in which you appear or are recommended. Because the set doesn't change, movement reflects your work rather than a new set of questions.
Measure the same questions, the same way, every month. Everything else is anecdote.
Want the baseline done for you?
The AI Visibility Audit runs 100 prompts across four engines, logs every source and hands you the gap map with a 90-day plan. Or start with the free 30-prompt scan.
