01 / Define
What exactly are we counting?
A ten-metric dictionary with units, formulas, evidence requirements, and a structured download.
AI visibility metrics: impressions, citations, referrals, and outcomes ↗A focused collection / Measurement
AI search measurement.
From exposure to evidence.
Understand where your content appears, whether an answer visibly credits your source, and which visits reach the site. Start with the question you need to answer.
02 / Observe
Is Google showing us in AI search?
Read the generative AI report, preserve its scope, and review pages with a filled worksheet.
How to read Search Console’s generative AI report ↗03 / Check attribution
Does the answer visibly credit your source?
Record citations, visible domain labels, author mentions and evidence in a repeatable observation ledger.
How to measure AI citations and visible brand attribution ↗04 / Follow the visit
Which assistant referrals reach the site?
Inspect collected GA4 acquisition, including the boundary between AI Assistant and Google Organic Search.
How to find referral traffic in GA4, including AI ↗05 / Reconcile
Why do the dashboards disagree?
Compare clicks, sessions, and page views with their collection rules and date boundaries intact.
Why Search Console, GA4, and Vercel show different traffic numbers ↗06 / Plan
What would a useful click target require?
Diagnose impressions with few clicks and test exposure scenarios in a local calculator.
Search impressions but no clicks: a diagnostic and planning calculator ↗07 / Evaluate
Does the evidence support a business decision?
Keep access, observed citations, referrals, and qualified outcomes separate in the review.
Measure AI visibility without confusing it with traffic. ↗Take it into your workflow
Useful beyond the article.
Every guide has matching Markdown and JSON. The downloads preserve assumptions, sources, and missing values, so a person or an agent can inspect the same evidence.
Revisit the question as evidence grows.
Provider reports change. Each guide records its source checks and editorial dates. The examples are illustrative; measured audience results require their own saved observations.