Google recently rebuilt its home page, replacing the familiar Search button with a set of AI features. An experiment by Productrise, a firm that does search optimisation for e-commerce, now reports that the same products shown in Google's AI Mode were on average 21.6% more expensive than in traditional search results, and that AI Mode product cards sat 49% higher on the page. Google said it had not verified the accuracy of the report, and that shopping results in both modes draw on the same data source, its Shopping Graph.
The finding lands in a live regulatory argument. Lawmakers are debating whether to intervene where companies use AI to analyse shoppers' personal data and adjust prices in real time — a practice known as surveillance pricing. Productrise's experiment does not demonstrate that. It shows nothing about Google setting different prices for different users based on what it knows about them, and the study does not claim to. What it does show is that two surfaces of the same search engine, fed by the same catalogue, put different products from different sellers at different prices in front of the same query.
The seller divergence is the more concrete result. When a product appeared in both modes, the primary seller differed 49.6% of the time. Productrise reads that as AI Mode favouring a different set of merchants, one less oriented toward low prices. Google's response — that both modes use the Shopping Graph — is true and beside the point. A shared input does not explain divergent output, and nothing in the company's answer addresses ranking or seller selection, which is where the difference actually lives.
Before the 21.6% figure gets quoted as a price premium, it is worth knowing what it was measured on. For the same query on the same day, only 1.28% of products appeared in both modes. The price comparison rests on that overlap. On any ordinary reading, two result sets sharing 1.28% of their items are not two views of one marketplace; they are two marketplaces, and comparing their prices is closer to comparing storefronts than to catching a retailer marking up the same item.
Independent search optimisation consultant Brodie Clark characterised the work as an analysis of the Shopping Graph and of how free product listings are placed differently in regular search and AI Mode — a description that is narrower than the price-gouging framing the headline number invites. Clark also suggested that a product's price counts for less in AI Mode's recommendations than one might expect, which is a more unsettling claim than the markup itself: it implies the ranking is optimising for something the shopper cannot see.
Whether shoppers actually pay more is a separate matter. The study found that many of them route around the AI feature and keep hunting for the best combination of price and offer. Clark's read is that the higher price can show up in the AI Mode grid and distort the first comparison a shopper makes, but that the click still tends to go to the lowest-price seller. Across queries as specific as "metal bidet attachment" and "stainless steel hand coffee grinder", Productrise found the same pattern: products recommended in AI Mode that also turned up in ordinary results were almost always dearer in AI Mode.
My reading is that the pricing angle is the weakest part of a genuinely useful study, and the placement numbers are the strongest. A card positioned 49% higher on the page is a distribution fact, not a statistical artefact of a thin overlap, and it applies to every product AI Mode surfaces rather than the 1.28% that matched. The thing to watch is not whether Google is overcharging — the evidence here does not support that — but that a ranking surface with near-total placement advantage is selecting sellers on criteria nobody outside Google can inspect. Google's rebuttal, that shoppers can open a product card and compare sellers themselves, quietly concedes the design: the AI layer is not the answer, it is a screen in front of the answer.
That is the tension with Google's own promise to make shopping simpler. The same dynamic has already appeared elsewhere in the product — AI Overviews reduce the clicks that reach publishers — and in shopping search, advertisers are finding it harder to control how their products are presented inside AI Mode. Senior search optimisation strategist Caitlin Geary put the consumer side of it plainly: AI search was supposed to spare shoppers from comparing prices across five open tabs. A system that defaults to the pricier option leaves that job exactly where it was, with one more layer on top of it.