Why this matters
AI shopping agents build shortlists from structured, current evidence. A useful measurement plan starts with stable questions and verifiable product facts.
Build a repeatable baseline
Choose a representative query set, record the product and channel state, and keep the evaluation inputs unchanged while you compare results.
Visibility and recommendation are different systems. Measure the outcome your team actually needs.
What to do next
Connect the catalog, teach the agent approved brand context, then monitor discovery and attributed sales together.