Setting up a generative AI search performance scorecard: what to track weekly, monthly, and quarterly

Key takeawaysn- AI search changes fast enough that a quarterly-only review will miss the story. The GPT-5.3 rollout in March 2026 cut average ChatGPT citations per response by roughly 27% in a single day, and Reddit's ChatGPT citation share collapsed ~86% overnight on August 14, 2026. Weekly tracking catches these events; annual audits don't.n- Split your scorecard into three cadences: weekly for operational signals (prompt visibility, crawler health, traffic blips), monthly for trends (citation share, content mix, conversions), quarterly for strategy (revenue attribution, prompt library refresh, competitive benchmarks).n- Visibility metrics alone are a vanity trap. Pair mention rate and citation rate with AI referral traffic and conversions, or you'll end up with a dashboard that looks good and proves nothing.n- Different engines need different treatment. Perplexity is stable enough to be your GEO test bench; Microsoft Copilot is volatile enough that weekly snapshots will mislead you.n- A scorecard is only worth what you do with it. Every weekly review should end with at least one action, even if it's just