Forty-five percent of marketing leaders admit they cannot accurately measure their brand's AI search visibility. Meanwhile the assistants they can't measure — ChatGPT, Gemini, Perplexity, Google AI Mode — are quietly deciding which brands get named when their buyers ask what to purchase. An AI visibility audit closes that gap: a structured, repeatable assessment of where you stand, why, and what to fix first.
Plenty of "audit" guides stop at screenshots of a dashboard. This is the GEOly AI version: seven steps, each with a concrete output, covering the layer most audits skip entirely — the commerce layer that decides whether AI agents can actually buy from you.
Key takeaways
- An AI visibility audit answers four questions in order: how visible are you (baseline), where exactly do you win and lose (prompts), who gets trusted instead (citations), and what blocks you (technical + commerce gates).
- Audit per engine, never in aggregate. Brand visibility fragments hard across models — a strong ChatGPT presence says nothing about Gemini.
- The two most skipped steps are factual-accuracy checking (what AI says about you can be wrong) and the commerce layer (feeds, checkout flags) — both are where real revenue hides.
- Every finding gets an evidence tag — confirmed, correlation, or hypothesis — so the roadmap isn't built on vibes.
- Output is a sequenced roadmap: technical unblocks in week one, content and entity fixes in month one, source-building across the quarter.
Before you start: fix the measurement frame
Two rules make the difference between an audit and a screenshot tour. First, measure against a fixed prompt portfolio — 100+ real buying questions across awareness, consideration and purchase, built the way our prompt tracking guide describes — not whatever queries came to mind that morning. Second, sample repeatedly: AI answers are probabilistic, so a single run per prompt produces an audit that won't replicate next week.
The seven steps
Step 1 — Baseline Share of Model, per engine
Run the portfolio across ChatGPT, Gemini, Perplexity and Google AI Mode; record mention rate, average position, and sentiment per engine — your Share of Model. Output: one table, engines as columns, you versus top competitors as rows. The per-engine split is usually the first surprise: the same brand routinely scores 3× higher on one engine than another.
