What should I use to measure answer-engine visibility by topic?

If you mean “How visible is our brand inside AI answers for each topic we care about?”, I’d measure it with a topic-level Share of Answer dashboard, not traditional keyword rankings.

Current AI-visibility measurement frameworks converge on tracking a fixed prompt set and measuring mentions, citations, recommendations, and competitor share separately. WeaveAI+1

The metric stack I’d use

Metric	What it tells you
Topic mention rate	% of prompts in a topic where your brand appears
Topic recommendation rate	% where the AI actually recommends you
Topic citation rate	% where your site/content is cited
Topic Share of Answer	Your share of brand mentions/recommendations vs competitors
Answer prominence	Whether you're #1 recommendation, shortlisted, or merely mentioned
Description accuracy	Whether the AI describes your product correctly
Citation source share	Which domains/pages are feeding the answers
Engine coverage	Visibility separately across ChatGPT, Gemini, Perplexity, Google AI Overviews, etc.

The important part is keeping the dimensions separate. A single “AI visibility score” can hide that you're highly visible for one topic but completely absent from another, or that you're frequently mentioned but rarely recommended. Cubed+1

For topic-level reporting

I'd structure your data roughly like:

Topic → Prompt → Engine → Run → Brand outcome

For example:

Topic: Project management software 50 buyer prompts → Mention rate: 42% → Recommendation rate: 28% → Citation rate: 21% → Share of Answer: 18% → Competitor A: 31% → Competitor B: 24%

Then break that down by informational / comparison / alternative / implementation / use-case prompts. This is much more actionable than “we have 63% AI visibility.” WeaveAI+1

What I'd actually use

You have two choices:

1. Build it yourself — best if you want maximum control. Maintain a fixed panel of ~50–150 high-value prompts per topic, run them repeatedly across the engines, and store the raw answers plus citations. Repeated sampling matters because AI answers are variable. WeaveAI+1

2. Use an AI-visibility platform — better if you need this at scale. Look for one that lets you slice results by topic + prompt + engine + competitor, rather than giving you only an overall AI score. Several current platforms explicitly support topic-level visibility and these dimensions. Mention Network+1

My recommendation: make Topic Share of Answer your headline KPI, with mention rate, recommendation rate, and citation rate underneath it. Keep the raw prompts/responses behind every number so you can diagnose why a topic is winning or losing.

If you tell me what you're measuring (e.g. SEO/AEO for a SaaS brand, agency reporting, or an internal dashboard), I can recommend the specific tools and exact KPI formula/dashboard structure I'd use.