Measure AI-associated inbound leads

What counts as a lead in this report?

Define a lead before choosing a channel. Specify the event that creates it, the qualification rule, the reporting period and how repeated contacts are deduplicated. A content download, a support request and a sales-qualified inquiry should not silently become the same outcome.

Use the existing CRM process where it fits. Add fields for observed acquisition evidence and buyer-reported research separately, without overwriting the organization’s established source history. Assign an owner to resolve duplicates and ambiguous cases. Do not claim that a CRM is incapable of recording an AI-related source; a properly designed field can capture what the buyer or tracking evidence establishes.

Which acquisition data is available?

In GA4, open Traffic acquisition and inspect the Session default channel grouping dimension. The current channel definitions include AI Assistant for recognized assistant referrers. Google AI Overviews and AI Mode remain part of Organic Search. Inspect Session source / medium to understand the recorded source.

Use the Traffic acquisition documentation to keep session scope consistent and distinguish configured key events from qualified leads. Review consent, implementation and referral limitations with the analytics owner. A report only describes the events and attribution it actually captured.

For campaign links you control, use a documented tagging convention and verify that it survives redirects. Google’s tagging guidance explains traffic-source dimensions. Do not invent campaign tags for links that an independent assistant may emit.

How do you collect buyer-reported evidence?

Ask a neutral question at an appropriate point: “What sources helped you evaluate us?” Allow several answers and an open response. Separate initial discovery from later comparison, because a buyer may encounter the brand in a referral and then consult an assistant.

Keep the buyer’s statement in the authorized CRM context and record the date. A statement such as “I checked your integration in ChatGPT” supports reported evaluation use. It does not necessarily identify the first touch or prove that the assistant caused the purchase. Avoid requesting private conversation content unless there is a specific, appropriate need.

How should overlapping evidence be counted?

Use categories that can overlap internally while deduplicating the overall lead total. An illustrative month contains 12 qualified leads with recorded assistant referrals and 8 with buyer-reported AI evaluation; 3 appear in both groups. The combined observed set contains 17 distinct leads, not 20. This is arithmetic for the fictional example, not a benchmark.

How does visibility measurement fit?

A fixed answer panel can show what the tested systems recommended during the period. It does not count the people who saw those answers. Keep recommendation observations beside the lead ledger as context, with separate denominators.

Report qualified leads, opportunities and revenue only where the CRM evidence supports those stages. Distinguish association from incrementality, and describe unresolved attribution honestly. This provides a useful commercial assessment without pretending to observe every earlier touch or guaranteeing a future lead volume.