Same Product, Same Day, 21.6% Pricier in Google's AI Mode

TL;DR

Productrise ran product-centric shopping queries through Google's AI Mode and traditional search on the same calendar day, every day from August 9 to August 31, 2026, and logged more than 2 million product listings across over 100,000 result pages. Two findings land hard on anyone who sells things online. First, the surfaces barely agree on what exists: only 1.28% of the products that rank in traditional search also appeared in AI Mode for the same query. Second, when they do agree on a product, the offer AI Mode leads with is 21.6% more expensive on average. Google's answer, given to MediaPost, is that both surfaces run on the same Shopping Graph and that it has not yet verified the study's accuracy.

One catalog, two very different shortlists

Both surfaces drink from the same well. Google's Shopping Graph holds more than 50 billion product listings and refreshes over 2 billion of them every hour. AI Mode layers Gemini on top of it, fans a shopping question out into several sub-queries, and returns a small panel of products instead of a carousel.

Small is the operative word. Productrise found traditional search returning 27.8 products per query against 3.9 in AI Mode, with 0.94 products in common. That is not a ranking shuffle. That is a different shelf.

This is the second time Productrise has measured the gap, and it is widening in the only direction that matters to a merchant. Its July study, run over 21 days on a comparable dataset, found traditional search returning products for 88% of tracked queries against roughly 23% for AI Mode, with a 0.8% overlap. August moved that to 1.28%. Both numbers round to nothing.

The price gap, measured three ways

The headline number is a matched comparison, which is the honest way to do it. Productrise pairs listings using Google's stable product identifier where one is present, restricts the match to the same query on the same calendar day, and compares lead offers, meaning the offer sitting in the top listing position on each surface. On those matched pairs, AI Mode ran 21.6% higher on average.

Zoom out to every listing tracked, matched or not, and the median tells the same story with a blunter instrument: $149 in AI Mode against $100 in traditional search, roughly 49% apart. That number mixes different products, so it measures the shelf rather than the price tag, but the shelf is what a shopper sees.

The third cut is the most useful one, because it separates disagreement from magnitude. On matched products, the two surfaces quote a different price 38.1% of the time. When they disagree, AI Mode is the more expensive one in 68.4% of cases, by a median of 22.2%. When AI Mode is cheaper, it is cheaper by a median of 7.8%. So the gap is not a wash with a fat tail. It is lopsided in both frequency and size.

Nobody is inflating anything

Here is the part worth getting right, because the cheap version of this story is wrong. Productrise does not claim Google marks prices up. The gap looks like a side effect of which seller gets picked.

On 49.6% of matched products, the two surfaces attribute the product to a different seller. Traditional search shows a popular products carousel that sorts merchants for the same item and rewards the lowest offer. AI Mode picks one offer per product card, and price is not the axis it optimizes.

Think of a shop assistant sent to fetch a specific item. Classic search checks every aisle carrying it and comes back with the cheapest box. AI Mode grabs the first box that satisfies your description, which is a perfectly reasonable thing to do right up until you learn the same box was two aisles over for 22% less.

That distinction matters legally as well as technically. What Productrise measured is the offer AI Mode selects, not a price personalized to an individual shopper. Those are different accusations, and only one of them is in this dataset.

Google's response

Google told MediaPost that "all shopping results on Google Search, including AI Mode and the search results page, are powered by the same data source, Google Shopping Graph," and added that it has not yet verified the accuracy of the report's claims. Both statements can be true alongside the finding. A shared catalog says nothing about which row of that catalog a ranker chooses to surface, which is precisely the thing being measured.

What to do if you sell things

• Stop assuming rank transfers. A top position in the shopping carousel buys you a 1.28% chance of appearing in AI Mode for the same query. Track the two surfaces separately or you are flying blind on one of them.
• Being the cheapest may not be the lever it was. If AI Mode is not sorting your category by price, the discount you cut to win the carousel does not buy you the AI Mode card.
• Check who Google thinks sells your product. With a seller mismatch on nearly half of matched products, a reseller or marketplace listing of your item can be the one that shows up.
• Feed quality is now the entry ticket. AI Mode has 3.9 slots to fill instead of 27.8. Attribute completeness is what makes you eligible for the shortlist at all.

The caveats, stated plainly

The study is US and UK only, the two markets where the Shopping Graph feeds AI Mode directly. The traditional-search side is limited to popular products carousel listings, while the AI Mode side counts every tracked product card, so the two samples are not built identically. Sessions were clean and non-personalized, which is the right control for measuring the surface but means the numbers describe a default shopper rather than yours. And Productrise says it plainly: this is a snapshot of a surface Google is still actively changing. The July run and the August run already disagree, which is the point.

The findings were independently reported by PPC Land, Relevant Audience and Futurism, with consistent figures across all of them.

Key Takeaways

• Across 2 million-plus listings tracked over 23 days in August, the identical product matched on the same query and day cost 21.6% more on average in Google AI Mode than in traditional search.
• Only 1.28% of products ranking in traditional search appeared in AI Mode for the same query, up from 0.8% in Productrise's July run. Visibility on one surface does not transfer to the other.
• AI Mode returns 3.9 products per query against 27.8 in traditional search, with 0.94 in common. The median AI Mode listing is $149 against $100.
• Prices disagree on 38.1% of matched products, and AI Mode is the pricier one in 68.4% of those cases by a median of 22.2%.
• The likely cause is seller selection, not markup: 49.6% of matched products are attributed to a different seller, and AI Mode does not appear to rank on lowest price the way the carousel does.
• Google says both surfaces run on the same Shopping Graph and that it has not verified the study. That answers where the data comes from, not which offer gets chosen.

Sources: Productrise data study, Productrise July study, PPC Land, MediaPost, Relevant Audience, Futurism, Google Shopping blog