Google's AI Mode and Search overlap on 1.28% of the products they show shoppers

The 21.6% premium is computed on identical items, and identical items are rare in this data. Multiply the 27.8 products an average search page returned by the 1.28% of them that also showed up in AI Mode and you get 0.36 shared listings a query, roughly one every three queries ^[2]. Run it the other way and about one product in eleven that AI Mode puts in front of a shopper also ranked in traditional search ^[8].

Inside that matched set the two surfaces mostly agreed on price. About 38.1% of matched items differed at all ^[5], which leaves 61.9% priced identically on both ^[6]. Where they did differ, AI Mode was the dearer option 68.4% of the time, at a median premium of 22.2% ^[6]. Multiply through: 26.1% of matched items cost more in AI Mode and 12.0% cost less ^[4][5].

The wider figure, a $149 median price in AI Mode against $100 in traditional search ^[7], compares the whole assortment each surface returned. It describes what a shopper meets on each surface, and the identical-item test is the one that controls for the product.

Google, responding to the findings, said both AI Mode and traditional search pull from the same Shopping Graph, and that shoppers can click through any listing to compare prices from multiple sellers ^[9]. Take that at face value and the divergence comes out of selection: one database, two rankers, one of which has to fit its answer into a chat reply. Even on the products both surfaces did show, the seller listed first changed about half the time ^[8].

Productrise did not report how the 100,000 responses split between the two surfaces. At 27.8 and 3.9 products each, a stated total above 2 million listings implies at least about two thirds of them were traditional search pages ^[7].

Two readings survive this. If AI Mode's short list skews toward bundles and brand-direct listings, the premium is assortment mix, and the 61.9% of matched items priced identically is consistent with that. If the ranker weights conversion history or merchant signals that correlate with price, the premium holds at the item level and 26.1% is a floor. Twenty-three days of August data cannot separate them ^[1]. In my view mix explains most of the gap, and the overlap number is the one a merchant has to plan around; a rerun matched on item identifier, showing AI Mode dearer on the same identifier most of the time, would show me wrong.