Google AI Mode picks different sellers than Shopping carousel

Search Engine Journal reports that Google's AI Mode surfaces markedly different products and sellers from the standard Shopping carousel, even when users run identical queries. The divergence is not a rounding error in an algorithm. It is a structural split in how Google allocates commercial visibility, and it is happening now, without fanfare.

The underlying data, cited by Search Engine Journal, shows AI Mode rarely returns the same products as the carousel sitting just below it on the same results page. When the two surfaces do agree on a product, they frequently disagree on the seller. Price differences follow. A shopper seeing AI Mode first encounters a different commercial reality from one who scrolls to the carousel, even though both asked Google the same question.

Two surfaces, two winner lists

The mechanism matters. Google's standard Shopping carousel is fed by a well-understood auction: merchants submit product feeds, bid on placement, and the carousel ranks them by a mix of bid, relevance, and feed quality. Brands and their agencies have spent years learning to optimise for it.

AI Mode does not appear to run on the same logic. It generates a conversational answer that incorporates product recommendations, and those recommendations are drawn from a different retrieval layer. The criteria governing which sellers appear in that layer are not yet public. What is clear is that winning the carousel does not mean winning AI Mode, and losing the carousel does not mean losing AI Mode.

For any brand with a B2C product line, that asymmetry is immediately consequential. For the industrial groups, financial services firms, and multilaterals that make up a significant share of enterprise marketing budgets, the lesson is one step removed but no less sharp: wherever an LLM-powered surface is involved, the selection logic differs from the legacy system sitting beside it. Brands that assume continuity between the two will misread their own visibility.

The commercial stakes for retail-adjacent categories are direct. A consumer electronics brand, a building materials supplier with a retail arm, or a financial services firm offering comparison-ready products each face a world in which AI Mode may be promoting a competitor that has never appeared in the carousel, or demoting a category leader that has spent heavily on feed optimisation. The budget allocation question becomes acute: spend to win the carousel, spend to appear in AI Mode, or accept that the two require separate strategies?

The honest answer is that no one has yet published a reliable method for systematically winning AI Mode placement. Google has not documented it. The data so far describes the gap but does not close it. That uncertainty is itself the signal. Brands accustomed to traceable, biddable systems are now operating in a surface where the selection logic is opaque and the results diverge from everything they have previously measured.

The practical consequence for large B2B-facing organisations is a governance problem as much as a marketing one. If the procurement officer at a major industrial group searches for a category of components and AI Mode returns a different shortlist from the standard results, which shortlist shapes the initial consideration set? LLM answers are increasingly where consideration begins, not where it ends. A brand absent from AI Mode's answer is absent from that first cut, regardless of how well its feed performs in the auction below.

Google is, in effect, running a second visibility system in parallel with the first. The brands that will lose ground are those that treat the two as one.