Measuring SEO and GEO requires several views of the same customer journey. A search impression, an AI mention, a cited URL, a website visit, and a qualified enquiry are not interchangeable. A useful report explains which event was observed and what it can support as a business decision.

Start by naming the action you want a reader to take: request a quote, complete a purchase, subscribe, or use a support resource successfully. Then work backwards to the discovery signals that may contribute to it. This prevents an attractive visibility chart from becoming the objective by itself.

Keep four measurement layers separate

  • Search visibility: impressions, clicks, query groups, landing pages, and relevant search appearance data available in your tools.
  • Observed AI visibility: mentions, citations, cited URLs, and factual accuracy within a recorded question sample.
  • Website behavior: measurable visits, landing pages, useful interactions, and completed actions.
  • Business quality: qualified enquiries, valid orders, revenue where appropriate, and whether the visitor's needs fit the offer.

Use a shared page or topic identifier to connect these layers without pretending they describe identical populations. Search Console, an analytics system, and a prompt-monitoring sheet can count different events under different rules.

Define an AI observation set you can repeat

Choose questions that reflect customer decisions, not just brand searches. Record the exact prompt, date, language, region, product or model shown, and search mode. Save the cited URLs and enough answer context to review whether a mention was accurate.

Define the denominator before calculating a rate. In a hypothetical test, 12 citations in 40 recorded answers means a 30 percent citation rate within that sample. It is not a 30 percent share of all AI answers or users. If next month's sample changes, report that change rather than drawing a seamless trend line.

Keep mention rate and citation rate separate. Also flag negative or incorrect mentions. Visibility that misstates your service area or recommends an incompatible product needs investigation even if the total count rises.

Audit referral attribution

Inspect the actual source and medium values your analytics system records for known AI referrals. Test a real clicked link where possible and note redirects, consent state, and whether a referrer survives. Do not classify all direct traffic as AI traffic.

Google Analytics distinguishes user, session, and event scopes in its traffic-source documentation. Choose dimensions that match the question. The source of a session and the channel that first acquired a user can tell different stories.

Keep a maintained mapping of verified referral domains, and document its limits. Some visits will remain unattributed; some users may read an answer and return later through another channel. Report measured AI referrals as an observable subset, not a complete count of influence.

In Google Analytics, also check the current default channel definitions. The AI Assistant channel covers recognized assistant sources, while Google AI Overviews and AI Mode are included in Organic Search. Keep those categories distinct when reporting results.

Build a small monthly scorecard

Use one row per important topic or landing-page group. Include a consistent date window and definitions alongside the figures:

  1. Non-branded organic clicks to the group, with the selected search filters.
  2. Observed citation count divided by recorded answers in the fixed question sample.
  3. Verified AI referral sessions reaching the group.
  4. Completed target actions and the number judged relevant by the business.
  5. Technical or editorial changes made during the period.

Add a short interpretation rather than combining everything into an unexplained “AI visibility score.” For example: “Citations increased in the comparison sample, but referral visits remain too few to evaluate conversion quality.” That is a useful finding even when it does not produce a dramatic headline.

Compare changes without claiming certainty

Save a baseline before revising a page. Record the deployed changes and compare similar time windows. Keep an unchanged page group as a reference when practical, while acknowledging differences in demand, competition, and seasonality.

If a guide gained traffic after adding a compatibility table, the sequence supports further investigation. It does not prove the table alone caused the increase. A concurrent promotion, indexing change, or seasonal demand shift may also matter. Google's search-traffic troubleshooting guide is useful when interpreting broader movements.

Small numbers need restraint. One additional order can move a conversion percentage sharply when the sample is tiny. Show counts next to rates and avoid declaring a winner from a handful of visits.

Turn the report into the next action

Use the combination of signals to choose work. Good impressions with weak clicks suggest a different investigation from inaccessible pages. Accurate citations with weak landing-page engagement may point to a mismatch between the answer and the destination. Useful traffic with irrelevant enquiries may expose unclear positioning.

The GEO audit guide helps connect those findings to technical and editorial tasks. Keep the reporting cadence stable, explain uncertainty, and evaluate whether the work improves the reader's path to a useful outcome.