wandered
Home/Resources/AI traffic/Checklist

AI traffic measurement checklist: from source rules to outcomes

Reliable AI traffic analytics is a chain. A clear channel label cannot compensate for a broken landing event, undefined conversion, mixed bot traffic, or an attribution rule no one can explain.

8 minute readUpdated Evidence verified
Working definition

An AI traffic measurement checklist verifies that source classification, collection, outcomes, privacy controls, machine separation, and reporting language work together before the metrics are used for decisions.

  • Validate collection with controlled visits and raw evidence.
  • Define meaningful actions and attribution rules before publishing rates.
  • Audit private data, retention, bot separation, and unknown coverage.
01

1. Verify acquisition collection

Test the handoff rather than trusting a channel label in isolation.

Confirm that the analytics tag or server event runs on every public landing page under the appropriate consent state. Inspect Session source, medium, channel, landing page, hostname, and redirects for controlled referral visits.

Document Google’s AI Assistants channel and the separate Organic Search treatment of AI Overviews and AI Mode. Preserve a versioned custom source list only where it adds a clear need.

  • Tag and consent behavior
  • Source / medium
  • Default channel group
  • Landing page and hostname
  • Redirect preservation
  • Internal and test-traffic exclusion
02

2. Verify the journey and outcomes

Every reported conversion needs an owner and a repeatable trigger.

Create an event dictionary for meaningful actions, trials, leads, purchases, and qualified pipeline. Test the source persistence, cross-domain flow, account continuity, value, deduplication, and attribution window.

Check that counts reconcile at the event level before calculating rates. Show raw counts when provider samples are small.

  • Event name and business meaning
  • Trigger and deduplication
  • Source continuity
  • Attribution window
  • Value source
  • QA owner
03

3. Verify separation, privacy, and language

A technically correct event can still produce a misleading report.

Exclude automated requests from human visits. Store crawler and agent evidence in a separate stream with verification confidence, applicable access policy, retention, and redaction.

Review every headline and summary for the evidence class it claims: observed, associated, self-reported, modeled, or unknown. Add collection-coverage notes and a last-reviewed date.

AuditPass condition
Human / bot separationAutomated requests never inflate visitor metrics
PrivacyPurpose, consent, retention, and access are documented
AttributionObserved and modeled values remain separate
CoverageUnknown and referrer loss are visible
MaintenanceProvider and channel rules have an owner

Methodology and verification.

Last verified August 17, 2026. The page is updated when the underlying analytics or provider documentation changes materially.

  1. 01

    Reviewed the linked primary documentation and separated provider claims from observations a website can verify.

  2. 02

    Kept human referrals, sampled answer visibility, machine requests, and modeled influence in separate evidence classes.

  3. 03

    Marked limitations wherever the available source or request data cannot support a provider-level conclusion.

Verify the evidence.

Provider behavior and analytics definitions change. These are the primary references reviewed for this page.

Questions teams ask.

How often should the checklist be run?+

Run it at implementation, after analytics or consent changes, after major site migrations, and whenever providers or GA4 classification rules change.

What is the first test?+

Verify a controlled recognized referral reaches the expected landing page with the correct session source and does not get overwritten by internal navigation.

Should bot logs be imported into GA4?+

Not as human sessions. Machine evidence can be analyzed separately and joined only in a model that preserves the actor and denominator.

Get one email when the product goes live.

One launch email. No account, no weekly drip, no noise.