Why quarterly?
Because the failure modes are slow. Robots.txt drifts after a platform update; dates fall out of sync page by page; a directory changes its fields and your profile goes stale; an engine ships a model update — Gemini 3 becoming default in January 2026, the AI Overviews and AI Mode merger at I/O 2026 — and your baseline quietly stops describing reality. Monthly measurement catches score movement; only a structured quarterly audit catches the causes. One week, five sections, fixed agenda, every quarter.
The template assumes one person can query logs, edit content and update profiles; if your permissions differ, split the days across owners but keep the single readout. Fragmented audits produce fragmented fixes, and the readout is where causes meet consequences.
Section 1: Technical access re-verification
- Re-read robots.txt and confirm OAI-SearchBot, ChatGPT-User and PerplexityBot access survived the quarter’s deploys. Re-confirm the GPTBot and Google-Extended decisions are still what you intend (training-only implications, not visibility).
- Check Bing Webmaster Tools health and IndexNow delivery; the Bing index still feeds ChatGPT’s search layer.
- Segment the quarter’s server logs by AI user agent: fetch volume trends, error rates on answer-bearing paths, redirect chains introduced by migrations. See AI crawler logs.
- Re-validate schema on a sample of templates; tag-manager injection breakage is the classic silent failure.
Section 2: Content decay check
- Audit dates on the top 30 citation pages: visible, meta and schema must agree.
- Re-verify every statistic older than six months; restate source and date or retire the claim.
- Re-run fan-out mapping on your five most valuable topics — engine sub-questions drift as models update, and last quarter’s brief may miss this quarter’s tree. See query fan-out.
- Flag pages where new competitor content now answers the question better; grounding compares.
Section 3: Source-graph drift
- Re-run 20-30 category questions and diff the cited-domain list against last quarter; new entrants are your new PR targets.
- Re-audit NAP and brand facts (name, founding date, address, description) across profiles; fix every drift.
- Review-flow health check: invitation volume steady, response rate 100%, no velocity anomalies that read as manipulation.
- Entity anchor check: Wikipedia/Wikidata status unchanged, sameAs set still complete. See the 30-minute entity audit.
Section 4: Measurement integrity
- Question-set hygiene: retire questions buyers no longer ask, add new ones from sales and support language — but version the set explicitly and keep the old version’s archive comparable.
- Confirm every monthly run happened on schedule and every score links to screenshots.
- Review the hallucination log: what errors appeared in answers, which were fixed at the source, which persist. See the measurement standard.
Section 5: The readout
One document, fixed format: score trends per engine across the four dimensions (Visibility, Citation Rate, Semantic Coverage, Brand Perception); the quarter’s shipped interventions mapped to observed deltas; the decay items found and fixed; and next quarter’s priorities chosen from the 50-point checklist group with the lowest score. Thirty minutes to present, five minutes of it spent on what did not work — the section that makes everything else believable.
What does the calendar look like?
| Day | Section | Output |
|---|---|---|
| Mon | Technical access | Crawler/index status sheet, fix list |
| Tue | Content decay | Refresh list with owners |
| Wed | Source graph | Drift fixes, new PR targets |
| Thu | Measurement integrity | Versioned question set, archive audit |
| Fri | Readout | One document, next quarter’s priorities |
Run four of these and the audit stops being an event and becomes the program’s immune system: decay gets caught in weeks instead of quarters, engine changes get absorbed deliberately instead of accidentally, and the annual strategy conversation starts from four readouts of evidence instead of one afternoon of opinions. That is the whole template — the discipline is the reusable part.