Arrow 
Measure each page against its intended decision, then keep visibility, visits, completed actions and qualified inquiries separate. Use a fixed page cohort, a stated time window and a clear denominator. A booking-button click is an interaction; a confirmed booking is a different event.
Define success before you publish
A comparison page should help a suitable buyer evaluate alternatives. An eligibility page should explain who can use the offer. A booking page should help an eligible person complete a request. Those jobs deserve different success checks.
Document one primary outcome for each cohort. Examples include confirmed trial starts, completed appointment requests and fewer unqualified inquiries. Pair the outcome with a diagnostic measure such as form errors so you can distinguish poor intent fit from a broken flow.
Separate an interaction from a business outcome
Google Analytics events record interactions. Your implementation must still define what each interaction means. A clicked submit button does not confirm that a form was accepted.
Count the success state after the server confirms receipt, and deduplicate repeated submissions where appropriate. Keep contact details in your authorized operational system. Use a non-identifying inquiry ID to connect a completed request with its later qualification status.
Write the denominator into every report
| Observation | Count | Interpretation |
|---|---|---|
| Tracked landing sessions | 200 | Observed sessions, subject to consent and tracking limits |
| Unique completed inquiries | 8 | Accepted requests after deduplication |
| Qualified inquiries | 3 | Met the pre-agreed fit criteria |
In this constructed example, completed inquiries per tracked landing session are 8 ÷ 200 = 4%. Qualification among inquiries is 3 ÷ 8 = 37.5%. Neither percentage is revenue, and neither establishes that an AI answer caused the inquiry.
Do not mix user, session and inquiry denominators under the same label. If qualification has not been reviewed yet, mark it pending rather than counting the record as unsuitable.
Use intent cohorts instead of one site-wide average
Group pages that support comparable decisions, then hold the membership stable for the reporting window. Keep comparison content, factual verification pages and action pages separate. Include the time period and source coverage in the report.
A small cohort can swing because of one inquiry. Show the counts beside rates, and avoid declaring a winner from a handful of observations. Annotate changes in pricing, campaigns and form behavior before attributing an improvement to content alone.
Keep attribution honest
A referrer can identify some observed visits, but it cannot reconstruct every answer a buyer saw. Ask an optional discovery question in your intake flow and preserve its answer separately from technical attribution.
Read the attribution guide for the wider measurement system. Use citation records for observed answer visibility. Together they describe different parts of the journey; they do not justify filling missing data with a success claim.
Sources and editorial scope
Published October 20, 2026. This is Arrow AI’s editorial method. Examples and scores are illustrative, not measured demand, customer results or AI ranking factors. The linked documentation provides platform context.
The Arrow read
Connect public visibility to accountable work
Arrow AI brings AI visibility and business workflows into the same conversation. Begin with the free GEO audit to inspect your public presence, then define how your team will recognize a completed request and a qualified inquiry. Keep those outcomes distinct from a visibility score. Run the free GEO audit.