A weekly AI search scorecard needs five numbers: citation rate on your priority question set, share of voice versus named competitors, description accuracy, AI-referred sessions from assistant browsers and answer links, and any movement in which sources the engines draw on. Everything else is monthly or quarterly; these five catch problems while they're cheap to fix.
The weekly scorecard
| Metric | How to measure | Alarm signal |
|---|---|---|
| Citation rate | % of priority questions where you're cited | Drop on questions you previously held |
| Share of voice | Your citations vs competitors, same questions | Competitor gaining on money questions |
| Description accuracy | Spot-check how engines describe you | New factual error or stale pricing |
| AI-referred sessions | Referrer + UTM analysis in analytics | Trend break after an engine update |
| Source movements | Domains cited alongside/instead of you | New source you have no presence on |
What stays monthly and quarterly
- Monthly: full question-set sweep across all engines, sentiment scoring, content refresh queue reprioritization.
- Monthly: correlation of AI-referred sessions with signups and pipeline.
- Quarterly: full visibility audit, competitor deep-dive, and question-set review (add new buyer questions, retire dead ones, versioned so trends survive).
