Google's product carousel is gone: your webshop is now second in line

Google's organic product carousel has all but vanished across the European Economic Area. Not weakened, not pushed further down the page. Gone. The Dutch tracking platform Productrise measured drops of 90 to 100 percent within days, while in that same measurement the United Kingdom, the United States and Australia carried on at their usual level.

What replaced it is not a smaller carousel. It is a different running order. Google now documents one block for comparison services and one block for direct providers, and Google's own documentation on the supplier unit contains the sentence that settles everything: the supplier unit only appears when the aggregator unit appears. In plain terms: in the new European product display, your brand's own shop has become dependent on a comparison site summoning the block above it.

One number is missing from everything below, and from Google's own documentation too: how many clicks this actually cost. Nobody has published it. What does exist sits in Google's developer documentation, in the European Commission decision and in the Productrise measurement, with limitations that are substantial here.

What exactly happened to the product carousel?

Google rolled out the new layout on 8 September. A week later, on 15 and 16 September, Productrise watched the "popular products" block collapse in five European countries at once. Productrise tracks commercial queries daily and measures, per country, the share in which at least one organic product carousel appears. That share collapsed to near zero within days in Germany, France, Belgium, the Netherlands and Sweden. In the United Kingdom, the United States and Australia the same measurement did not move.

That control group is the strongest part of the evidence. A general algorithm update does not hit five countries inside one region and leave three countries outside it untouched. A regional rebuild does exactly that.

The size of the fall is on record too. Historically, according to Productrise, roughly 85 percent of the commercial queries they track triggered at least one organic product carousel. This was not a niche feature quietly retiring. It was the default picture at the top of a product search.

Now the limitations, because I am not going to bury them at the bottom. Productrise sells rank tracking for precisely these carousels. The company writes in the same publication that its own product value in the EEA has been sharply reduced for the time being, which pushes the conflict of interest away from exaggeration rather than towards it, but it remains a single vendor with its own dataset and no replication by a second party.

More importantly: this is a measurement of search results, not of clicks. Nobody has published this month how much organic traffic actually disappeared. Anyone handing you a number about lost revenue right now is extrapolating from something nobody has measured yet. Look instead at your own organic clicks in Merchant Center and in Search Console across the second half of September. That is the only measurement that is about your brand.

Why only in Europe and not in the United Kingdom?

Because the trigger is legal. The Digital Markets Act is Regulation (EU) 2022/1925 and applies across the European Union; it has not been incorporated into the EEA Agreement. Google's rebuild is visible across the whole European Economic Area because that is the rollout area Google chose, not because the law reaches that far. On 23 July 2026 the European Commission adopted two decisions finding that Google breaches the DMA, and imposed a fine of 460 million euros for favouring its own services inside Google Search. Another 430 million euros followed for obstructing app developers on Google Play, 890 million euros in total.

The obligation the Commission considers breached is stated briefly and without wiggle room: gatekeepers may not treat their own services more favourably in ranking than those of third parties, and must apply transparent, fair and non-discriminatory conditions to that ranking. The Commission explicitly names shopping, hotels, transport and sports results as the favoured own services, and ordered Google to bring the infringement to an end.

On 8 September Google rolled out the revised results in the EEA and documented it the same day. The Search Central changelog carries the entry on regional differences in the search experience, covering aggregator units, supplier units and carousels. That changelog is the moment this went from rumour to documentation.

Google itself is remarkably candid about how it rates the outcome. Speaking to Reuters, the company called the changes the worst degradation of service quality in the search engine's 29-year history. That is a party condemning its own remedy, so read it as a position and not as a fact. It does say something about direction, though: this is not a product improvement that happens to also satisfy the rules.

One detail I dropped because I could not find it in the primary source: a sixty-day compliance deadline is doing the rounds. It is not on the Commission's page. What is on that page is that Google was ordered to end the infringement. Anyone who needs the exact date should be reading the full decision, not the press release.

The sentence in Google's documentation that sets the running order

This is where the story sits. Not in the drop, but in what was rebuilt in its place.

For the EEA, Google documents two units side by side. It describes the aggregator unit as a multi-provider feature for Vertical Search Services, explicitly naming online travel agencies, Comparison Shopping Services, metasearch engines and directories. Inside that block, admitted aggregators fill in their own results, the results of the highest-ranked provider are expanded by default, and clicks lead straight to that aggregator's site. Only one appears at a time.

Next to it sits the supplier unit, for direct providers. Google gives individual hotels, airlines, retailers with a physical store and service providers such as plumbers as examples. For a brand with its own webshop, this is where you belong. The good news is in there too: you do not have to supply any data beyond what is reachable by crawling. No feed, no approval procedure, no application form.

And then the sentence this article is built around. Google writes that the supplier unit only appears when the aggregator unit appears.

Read that twice. The party without a gate has been made dependent on the party with one. Because the aggregator unit does have a threshold: you must be approved as a Vertical Search Service, you must supply the required data through direct feeds or a real-time API, and for product queries the application runs through the Comparison Shopping Services programme. That is precisely the layer the self-preferencing complaint came from in the first place.

So the outcome of this remedy is that access to the European product display has been delegated. Not to the seller, but to an intermediary layer that runs its own admissions process. Anyone who had organised their visibility entirely through Merchant Center lost their spot in that display this month without doing anything wrong. That is exactly the pattern we described when brands failed on marketplaces over rules they had not written.

What replaced the carousel?

In practice, European searchers now see a block Google labels as comparison sites. Several comparison services stacked underneath each other, each in its own expandable row, with one provider open by default. For a query like air fryer, Productrise names Redbrain, Shoparize and Bigshopper among others. Click inside it and you land on that comparison site's product page, not in a Google product view.

Google itself is listed as one of those comparison sites. Expand Google and you get something that closely resembles the old carousel, knowledge panel with multiple sellers included. It just sits one or two clicks deeper now, and the tiles no longer show a seller name on the card. You see a price and a pointer to more prices.

That last detail is the painful one for a brand. The shop's name has disappeared from the card. A buyer sees a product and a price, and only learns who is selling it after a click.

An honest footnote to that table: this is the picture as of mid-September. Productrise itself writes that the current arrangement is probably a starting point rather than an end state, and the Commission is still assessing the implementation. Build a three-year plan on this picture and you are planning around a screenshot. What you can do is read the direction, and the direction is clear: away from the feed, back towards your own site structure.

Is the structured data carousel on your category page the way out?

It is the route closest to your own site, and Google documented it this month in the same round. Structured data carousels are a rich result in which visitors scroll horizontally through entities from a single site. Google also calls this a host carousel. Available in the EEA, Turkey and South Africa, and in the EEA for product queries among others.

The technical requirement is concrete. You add ItemList markup in combination with at least one of the supported types: Product, LocalBusiness or Event. And then the clause that shifts the whole subject: that markup has to sit on a summary or category page, a listing-style page with information about at least three entities, which then links out to other pages on your own site for more detail about those entities. You do not have to mark up the detail pages themselves, but the URLs of those detail pages do have to appear in the markup on your listing page.

Read what is happening there. The surface a brand appears on moves from the product feed to the category page. That is not a feed-management job. That is a question about information architecture: which categories does this brand claim, what are they called, and do they link out cleanly. The seven elements we worked out earlier for GEO-ready category pages just got harder than they were.

Two things you need to know before you sell this as the rescue.

It is beta. Google says itself that requirements and guidelines may still change. There are no volume figures, no guarantee and no rollout schedule, and there is no public measurement this month of how often those host carousels actually appear.

And this one I only found by reading the page all the way through: on that same documentation page, Google points product queries in CSS programme countries towards the Comparison Shopping Services programme. So the gate-free route is less gate-free than it looks at first glance. For hotels, local businesses and things to do there is an ordinary interest form; for products there is a programme. Anyone claiming a bit of schema markup puts you straight back on the page has not read that line.

Seven months ago Amazon grew almost fifty percent on the surface that is now gone

Here is the reversal that finishes this story for me, and it comes from the same tracker.

In February 2026 Ecommerce News published, based on Productrise data, that Amazon had grown almost fifty percent in organic visibility on Google's first page in the Netherlands, while Bol showed a small decline. The measurement basis was stated too, from Hugo Huijer, the tracker's creator: every week they follow hundreds of thousands of organic product listings on Google's first results page. What is not stated is which queries those are, or how that sample is put together. That is where the weakness sits, not in the size.

Where did that growth sit? In the product carousels at the top and to the side of the page, on generic terms like wireless earbuds, laptop bag and gaming mouse. Huijer attributed the win to strategic feed optimisation, growing review volumes and a platform model that fits how Google ranks products.

Seven months later, that same tracker measures that the surface on which the win was booked no longer exists.

Note what I am not saying here. I am not saying the February measurement was wrong. The measurement was correct and the win was real. I am saying the playing field slid out from under the winner, by decision, within seven months, without the winner doing anything wrong. That is the difference between scoring on a position and owning one.

And that difference is not an abstraction. Feed optimisation is real work with a real return. It is just that the return depends entirely on the continued existence of the surface it gets paid out on. Put that next to what a marketplace takes in commission from your margin and you see the same pattern from another angle: the terms of the channel are not yours.

What you can do on Monday

No panic, but there is work. In this order.

Measure first, before you change anything. Put your organic clicks in Merchant Center and your Search Console data for the first half of September next to the second half. Split EEA from non-EEA if you sell internationally, because that is where the fault line runs. Without that baseline you will not know in three months what this change cost you and what had other causes.

Then walk through your category pages. The requirement is explicit: a listing-style page with information about at least three entities, linking out to individual detail pages on the same domain. Anchor links within the same page do not count. Check for each important category whether the page exists, whether it has substance rather than a grid of images, and whether the detail page URLs are present in the markup.

Add ItemList in combination with Product. Validate with Google's Rich Results Test, publish a handful of pages, and check with URL inspection how Google sees them. This is beta, so treat it as an experiment with a measurement plan, not as an implementation with a delivery date.

Leave your feed alone. The Merchant Center feed still powers paid campaigns and the Shopping Graph itself. This is not a replacement but a second system. Cut your feed work now because the carousel is gone and you damage the one channel that still pays out. How that feed carries through into AI recommendations we covered separately before.

And the most important one: add up how much of your demand arrives through a platform. Not your revenue per channel. Your demand. How many people search for your brand name, how many come direct, how many come from your own list. That percentage is the only part of your visibility that cannot be removed by a third party's decision.

What we changed in our own work this month

Since this month an e-commerce project with us starts with a category page audit, not with a feed check.

In six years we have worked with more than 200 brands. The ones hit hardest this month are the ones whose demand starts with a generic product term rather than with their own name. That was a rational choice for years. The channel worked, it was measurable, and it scaled faster than building a brand. What this month showed is the price of that choice, and that price gets collected in one go rather than in instalments.

Since September the first question in our e-commerce calls has changed. No longer how we rank in Shopping, but which part of the visibility sits on rented ground. A brand that only exists at the moment someone types a generic product term exists at the mercy of that results page's layout. A brand people search for by name still exists when that page changes. That is the same reasoning underneath our approach to brand positioning for AI agents and underneath what we wrote about agentic search and webshop traffic: the interface changes, the brand demand stays.

Concretely, from this month on we do two things differently in our e-commerce work. We include a category page audit as the first answer to the European rebuild, with ItemList and Product markup, the three-entity requirement and clean internal linking. And we say plainly that this is a beta whose reach nobody knows. What we do not do is sell it as a replacement for Shopping, because it is not.

Positioning that rests on a platform feature is rented ground. You can build on it perfectly well, as long as you know you are renting. What we do is grow the part you own: a brand people ask for by name, a site that gets found on its own, and a list that belongs to you. That is the same ground building an e-commerce brand stands on. Want to know which part of your visibility is rented? Get in touch.