Oli Yeates · 19 September 2026
Google's Rethink Retail 2026: what the first AI-powered golden quarter actually asks of you
Google streamed its Rethink Retail keynote on 16 September, fresh off the event in New York, and the framing was blunt: this is the first golden quarter where AI sits between your products and your customers at every stage of the journey.
Most event recaps will hand you a feature list. That is not the useful part. The useful part is that Google has quietly changed what an ad is, what your product feed is for, and which team in your business owns the bidding strategy. Here is what matters, and what we think UK retailers should do before Black Friday.
1. Ads are becoming answers, not creative
The line that ran through the keynote was that the best ads in the age of AI are answers. In practice, that means shopping results increasingly open with an AI-generated brief explaining what to consider when buying that type of product, followed by recommendations drawn from reviews, pricing and other organic signals rather than ad spend.
The implication is uncomfortable for brands who have invested heavily in polished creative and lightly in product data. If the system is assembling an answer, the brand that has supplied the most complete, accurate, structured information about its products is the one that gets assembled into it. Your feed attributes are now your ad copy.

2. Merchant Center has stopped being admin
Google pushed retailers towards submitting conversational and video attributes, and towards connecting loyalty member data so that member pricing and perks can surface in AI-powered shopping. Alongside that, AI Max for Shopping is now in open beta, with text customisation and Final URL expansion designed to catch evolving search intent automatically.
Both of those features run on feed depth. Material, fit, pattern, warranty, care instructions, returns policy: the fields most retailers leave half-complete are now reach drivers. Google's own claim is that merchants who adopt core feed practices see roughly a 5% conversion increase the following month. Treat the feed as a live optimisation surface with an owner and a cadence, not a setup task someone completed in 2023.

3. Bidding to margin, not to ROAS
The most significant commercial change is easy to miss. Product Value Optimization, announced as a beta, lets advertisers apply value adjustments so automated bidding pushes the products that matter to the business rather than simply the products that convert most cheaply.
That is a genuine shift in who owns performance marketing. Push high-margin lines harder, discount the weighting on loss leaders, clear overstock on purpose. For years the agency conversation has been about ROAS because ROAS was the only lever available. This moves the conversation to contribution margin, and it means your finance director needs to be in the room when the Q4 strategy is set. Most retailers we speak to do not have margin data structured at SKU level in a form that can feed this. That is the work to do now.
4. Shoppers are delegating research, so you have to win twice
Google cited Ipsos research suggesting that 99% of shoppers who research using ChatGPT still consult Google before making a final decision. Read that carefully, because it cuts both ways. Google is reassuring advertisers that it still owns the transaction. It is also confirming that a large share of the consideration phase now happens somewhere Google cannot sell you a placement.
The practical answer is that generative visibility and traditional SEO are increasingly the same discipline. Clear, factual, well-structured content about your products, credible reviews, transparent pricing and machine-readable specifications are what get you cited in an AI answer and what get you recommended in a shopping brief. There is no separate GEO budget line. There is content quality and data quality, doing double duty.

5. YouTube is being repositioned as demand capture
Google presented TransUnion marketing mix modelling showing Demand Gen delivering an 11% lift in online sales and 14% in-store, with higher ROAS than social publishers, and noted that 68% of Demand Gen conversions came from people who had not seen the brand's Search ads in the previous 30 days.

The bigger structural news is Business Agent. A Gemini-powered assistant sits inside the ad unit itself, answering product questions, comparing options and shortening the distance between inspiration and checkout. If that works at scale, YouTube stops being an awareness line on the plan and starts competing directly with paid social for lower-funnel budget.
The UK footnote Google did not dwell on
Worth flagging, because it affects planning. AI performance insights, which show brand share of voice across AI Mode and AI Overviews, launched in Australia, Canada, India, New Zealand and the US, while Universal Commerce Protocol cart transfer and checkout testing began in the US with Australia and Canada following early next year.
The UK is not in the first wave on several of these. That is not a reason to wait. It is a reason to do the unglamorous groundwork now, because when these surfaces open here the winners will be the retailers whose feed, content and measurement were already in shape. Readiness is the only part of this you control.
What we would do in the next thirty days
- Audit feed completeness by attribute, not by error count. Anything below 80% completion on descriptive attributes is a visibility problem, not a data hygiene one.
- Get margin data to SKU level. Even a rough three-tier banding is enough to start testing value-based bidding when it becomes available to you.
- Test AI Max for Shopping on a contained product set rather than across the account, and hold a control.
- Review your product content pages as answers. If a shopper asked an assistant three questions about this product, does your page contain the answers in plain language?
- Move a slice of paid social budget into Demand Gen and measure it properly, with incrementality rather than platform-reported conversions.
The honest read
None of this is optional automation you can decline. Google is consolidating discovery, recommendation and checkout into surfaces where the entry ticket is structured data rather than budget. The retailers who spent the last two years treating product data as an IT problem are about to find out it was a marketing asset.

If you would like a second opinion on how ready your feed and content genuinely are ahead of peak, that is precisely the kind of audit our paid media and SEO teams run together rather than separately.