How to measure Share of Answer, and what a good number looks like

Share of Answer is the percentage of tracked prompts where an AI engine cites or recommends your brand. It is the closest thing AEO has to a market-share metric, and it is easy to get wrong. A number built on the wrong prompts, or scored inconsistently, will move for reasons that have nothing to do with your work.

Start with the prompt set, not the dashboard

Your Share of Answer is only as meaningful as the prompts behind it. We recommend 150–300 prompts per product line, written the way buyers actually ask and grouped by intent. “Best payroll software for a 40-person startup” tells you far more than “payroll software.”

• Category prompts: “best X for Y” and “alternatives to X” questions, where shortlists get formed

• Problem prompts: questions that describe a pain without naming a category

• Comparison prompts: head-to-head questions that name you and a competitor

• Brand prompts: questions about your pricing, security or fit, where accuracy matters more than presence

Score citations and recommendations separately

An answer can include you as a linked citation, a named recommendation, or both. A citation means the engine used your page as a source; a recommendation means it put you in front of the buyer. They respond to different work, so score them as separate fields and run each prompt at least three times per engine per week, since a single run overstates both wins and losses.

What a good number looks like

It depends on how crowded the category is. Across the B2B software categories we track, the median brand appears in 14% of category-prompt answers, and the category leader usually sits between 35% and 50%. On brand prompts, anything under 90% tends to signal an accuracy or sourcing problem worth fixing first.

Plan by engine, report the blend

Engines diverge more than most teams expect. It’s common to see a brand at 30% on Perplexity and 8% on Google AI Overviews for the same prompts, because the answers draw on different sources. Report the blended number to leadership, but plan the work against the per-engine view.

1. Set a four-week baseline before changing anything.

2. Pick one engine and one prompt group to move first.

3. Check Answer Diff weekly to tie movement in the number to specific changes in the answers.

4. Log every content change so you can explain each point of movement.

Treat Share of Answer like any other share metric: a stable definition, a fixed denominator and a written record of what changed. Teams that work this way can defend the number in a planning meeting, which is what makes it worth tracking.