Most teams still measure search with rankings and clicks, which say little about whether AI answers mention you. AI-search measurement needs its own KPIs. Here's the set that matters and how to operationalize it.
The core KPIs
| Metric | What it tells you |
|---|---|
| Citation rate | How often you're referenced for target questions |
| Share of voice | Your citations vs competitors on the same questions |
| Sentiment | Whether you're described positively, neutrally or poorly |
| Source attribution | Which sources drive your appearances |
| Answer accuracy | Whether the model states your facts correctly |
How to collect them
- Define a question set that mirrors your funnel (awareness to decision).
- Query each model on a schedule and log citations, mentions and sentiment.
- Use a GEO tracking tool for scale, or a lightweight internal script to start.
- Segment by model and geography, results diverge across both.
