How much revenue is AI search worth to your catalog?

How the calculation works

Three layers, each visible in the breakdown panel:

The output is the incremental annual revenue captured with optimized PDPs versus leaving them in their current state. It is an estimate for sizing a decision, not a forecast: the model assumes your conversion rate holds for AI-referred visitors, which early data suggests is conservative — buyers arriving from an AI recommendation tend to land with the comparison already done.

Where the defaults come from

The 35% lift default derives from controlled comparisons of optimized versus unoptimized PDPs on the same SKU — pages that went from missing structured data and blocked AI crawlers to complete schema and open access. Specific high-intent queries can move 50–70%; the default averages down across a catalog where many SKUs are already partly optimized. The attention mechanics behind attribute-level matching are documented in Amazon Science's COSMO paper (SIGMOD 2024), which describes how Amazon's shopping AI maps commonsense buyer intents to products — the same class of matching the optimization work targets. For the framework behind the traffic-share inputs, see the What is GEO guide and the AI traffic analytics glossary entry.

What to do with the number

Three common uses. Budget case: pair the estimate with the prove-AI-search-ROI walkthrough for the board deck. Prioritization: if the upside is small because AI traffic share is already high, your problem is conversion, not visibility — audit PDPs with the AEO Grader instead. Reality check: run the SKU Visibility Grader to see whether the optimized state the model assumes is close or far from where your store is today. The gap between the two is the work; pricing tells you what closing it costs per month.