The news : On 1 September 2026, Productrise published a study comparing the products Google shows in classic search with those it shows in AI Mode. Across more than 2 million product listings tracked from 9 to 31 August 2026 in the United States and the United Kingdom, spanning more than 100,000 results pages and AI Mode responses, only 1.28% of products ranking in the Shopping carousel also appeared in AI Mode for the same query on the same day (Productrise, 1 September 2026).
Press coverage latched onto a different number: in AI Mode, the same products show up 21.6% more expensive on average. That is the figure that makes headlines, and Futurism turned it into a story about shoppers being ripped off. It is not the one that should worry you if you sell online.
The real lesson fits in one sentence: ranking well in Google no longer guarantees showing up in AI Mode. The two surfaces do not select the same products. And the gap is far wider than most brands assume.
What the study actually measured
The method is simple, and that is its strength. For twenty-three days, Productrise ran the same product queries against two Google surfaces: the Shopping carousel in classic search on one side, AI Mode responses on the other. Then it looked, day by day, for the products present on both.
The first finding is structural. AI Mode shows an average of 3.9 products per query, against 27.8 in classic search. The shelf is seven times shorter. Across every product tracked, AI Mode accounts for just 12.3%.
The second finding concerns the few products common to both surfaces. When the same product appears on both sides, the price differs 38.1% of the time.
And when it differs, it is almost always AI Mode showing the higher price: 68.4% of the time. The median gap then reaches 22.2% higher. When AI Mode is cheaper, the gap is only 7.8%.
One last figure, more unsettling still if you are a brand: in 49.6% of cases, the featured seller is not the same from one surface to the other.
Stop comparing matched products and compare the two catalogues as a whole, and the gap widens: a median price of $149 in AI Mode against $100 in classic search, roughly 49% more.
The number that matters is not 21.6%, it is 1.28%
This is where the study stops being a consumer story and becomes your problem. A 1.28% overlap means your position in classic search predicts almost nothing about your presence in AI Mode. These are not two displays of the same ranking. They are two different selections, made on the same day, for the same query, from the same declared catalogue.
Plenty of online retailers still steer their AI visibility on the opposite assumption: "I rank well in Shopping, so I will eventually show up in generative answers." That reasoning comes from a decade of SEO in which a single ranking scale governed every surface. It no longer holds. With 3.9 slots instead of 27.8, the step up is far higher, and being eleventh in the carousel is no longer a consolation prize: there is no eleventh place in AI Mode.
It is the same shift we described around the share of shopping queries triggering an AI Overview, and it follows the logic of zero-click search: visibility now plays out inside the answer, across a very small number of slots.
Do your products show up in AI answers?
We measure your presence across ChatGPT, Google AI Overviews, AI Mode and Perplexity, and tell you exactly what to fix. Free, no strings attached.
Half the time, the seller shown is not the one you think
The 49.6% different-seller figure deserves a pause, because it hits brands that sell direct while also being distributed through resellers. If your product surfaces in AI Mode but the featured merchant is a reseller, you have won the visibility and lost the transaction. You are funding a competitor's reputation on your own SKU.
Search Engine Journal raises an interesting hypothesis here: if AI Mode does not sort by lowest price first, unlike the classic carousel, then a brand unable to match the price floor might actually gain. That is a hypothesis, not a result: the study says nothing about the selection mechanism. We cite it because it is plausible and testable, not because it has been demonstrated.
What Google says
Asked by Futurism on 2 September 2026, Google disputes the reading of the study. Its position: all shopping results, whatever the surface, "are powered by the same data source: our Shopping Graph," and shoppers "can easily click into a product listing to compare prices."
That answer is accurate and does not address the question asked. Sharing a data source does not imply selecting the same items: the gap between the available catalogue and the displayed selection is precisely what the study measures. Google denies none of the figures; it restates an architecture. For a search practitioner the practical conclusion is unchanged: same feed in, two different outputs, therefore two surfaces to measure separately.
The limits of this study
Four caveats, to be stated before anyone builds a strategy on this.
The author is not neutral. Productrise sells a product-listing monitoring tool. In other words, a study concluding that product visibility has become hard to track serves its business directly. That does not invalidate the method: it is described and reproducible. But a result of this magnitude calls for independent replication, and nobody has published one to date.
The scope is American and British. Nothing in this data describes the French or wider European market. Google's surfaces roll out at different speeds by country and language. We have not measured an equivalent overlap ourselves. Transposing these percentages as-is to a European catalogue would be extrapolation, not analysis.
This is a shopping study, not a content study. The 1.28% overlap covers product listings, not editorial pages. Do not conclude that 98.7% of your well-ranked blog posts are absent from generative answers. Those are distinct selection mechanisms, and that measurement has not been made.
Twenty-three days in August is not a year. The window is short and seasonal. It is enough to establish a large gap; it is not enough to establish a trend, or to tell a durable state from a temporary setting.
One point of vocabulary matters here: saying AI Mode "prefers expensive products" is an interpretation, not an observation. What is observed is a correlation between the surface and the price level displayed. The mechanism remains unknown, and Google does not document it.
What to do this week
- Measure AI Mode separately from the Shopping carousel. If your reporting aggregates the two, it produces an average that describes neither surface. That is the direct consequence of the 1.28%.
- Check which merchant is shown on your own SKUs. Querying your ten most profitable products is enough to find out whether you are handing your sales to a reseller. It is free and takes twenty minutes.
- Work on feed completeness rather than price. Since AI Mode does not appear to sort by lowest price first, racing to the price floor does not produce the expected effect there. A rich, current feed with complete attributes remains the only documented lever; we noted the same around AI share of voice in Merchant Center.
To instrument all of this over time, Google's own reported data stays more stable than externally reconstructed sampling: see the AI visibility report in Search Console.
The Cicero take
This study does not show that Google is ripping shoppers off. It shows something more useful to anyone working on visibility: classic search and AI Mode are no longer two views of the same ranking. Seven times fewer slots, 1.28% overlap, a different seller half the time. At that scale of divergence, "ranking well" stops being an answer: the question becomes ranking well where. Brands that keep steering by a single curve will discover the second surface on the day they have already vanished from it.
If you do only one thing after reading this: take your ten best-selling SKUs, search for them in AI Mode, and note who shows up in your place. Twenty minutes, no tooling, and you will know where you actually stand.
Sources
- → Productrise, data study of 1 September 2026: more than 2 million product listings, more than 100,000 results pages and AI Mode responses, 9 to 31 August 2026, US and UK. 1.28% overlap, 21.6% price gap on matched products, 3.9 products per AI Mode response against 27.8 in classic search, 49.6% different main sellers, median price of $149 against $100.
- → Futurism, 2 September 2026: Google's official response, stating that all shopping results are powered by the same data source, the Shopping Graph.
- → Search Engine Journal, 2 September 2026: SEO reading of the study, hypothesis on brands unable to match the lowest price.
- → PPC Land, 2 September 2026: methodology detail and median gaps (22.2% higher, 7.8% lower).
- → MediaPost, 2 September 2026: coverage and perspective for advertisers.
Frequently asked questions
Does ranking well in the Shopping carousel guarantee appearing in AI Mode?
No. That is the central finding of the Productrise study of 1 September 2026: for the same query on the same day, only 1.28% of products ranking in the Shopping carousel also appeared in AI Mode. AI Mode also shows 3.9 products per response against 27.8 in classic search. The two surfaces must therefore be measured separately: a good position in one does not predict presence in the other.
Why does AI Mode show higher prices?
Nobody knows for certain, and that is worth saying plainly. The study finds that on products present on both sides, the price differs 38.1% of the time and AI Mode is the pricier one 68.4% of the time, 21.6% higher on average. But it does not establish the selection mechanism. Google, asked by Futurism on 2 September 2026, responds that all shopping surfaces draw on the same data source, the Shopping Graph. Saying AI Mode prefers expensive products remains an interpretation, not an observation.
Do these figures apply outside the US and UK?
Nothing supports that claim. The study covers the United States and the United Kingdom over a twenty-three-day window in August 2026. Google's generative surfaces roll out at different speeds by country and language. The percentages should not be transposed as-is to another market; the methodological principle, however, measuring AI Mode separately from classic search, applies everywhere.
Does this mean 98.7% of my content is invisible to AI?
No, and this is a common confusion. The study measures product listings in a shopping surface, not editorial pages. The mechanisms selecting a blog post for a generative answer differ from those selecting a product in AI Mode, and that measurement has not been carried out. The 1.28% figure should not be generalised to editorial content.
Growth and SEO & GEO content strategist, I founded Cicéro to help businesses build lasting organic visibility : on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.
LinkedIn