On 15 October, Google switched on Universal Cart for US shoppers, and most of the coverage filed it under "nice little holiday feature", somewhere between the gift guides and the annual debate about when it's acceptable to play Christmas music.

Fair enough. It is nice. One basket that follows shoppers across Search, AI Mode and Gemini, holds products from several retailers, and quietly keeps watch. Price drops? Ping. Item back in stock? Ping. Promo code floating about? Ping. And because it sits on Google Wallet, it'll even suggest which card earns the best perks (finally, a purpose for the card we all opened purely for the lounge access).

Shoppers pay inside the cart with Google Pay, or get sent to the retailer's site to finish. Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair and a long tail of Shopify merchants have signed up to one route or the other. (Notice who isn't on that list. Don't worry, Amazon gets its own section. Amazon always gets its own section.)

For twenty years, Google's commercial job ended at the click. It pointed, you went, the retailer did the rest. Universal Cart moves the basket, the price-watching and, optionally, the payment onto Google's side of that click.

If you sell on Amazon, this should feel oddly familiar. A bit like hearing your own voice on a recording.

Universal Cart runs on the Universal Commerce Protocol (UCP), the open standard Google introduced in January so AI systems can read catalogues and complete purchases. Merchants opt in through a UCP hub inside Merchant Center, with Australia and Canada pencilled in for next year.

There are two roads to a sale:

That transfer is, per Google's own developer documentation, one-way in this first release. The merchant gets a new cart. Nothing flows back. Change your price afterwards and Google's cart has no idea. (Less a handshake, more a parcel left on the doorstep while you were in the shower.)

On both roads, Google says the merchant remains the seller of record. You take the order and the liability. Reassuring! It also says nothing about who sees the shopper, who gets to market to them afterwards, or who keeps the record of what they almost bought.

Measurement is where this stops being philosophical. Merchant Center's UCP dashboard tracks buy-button clicks through to purchases and cancellations. The API reporting (still in alpha, naturally) adds cart transfers and average cart value at transfer. What none of the published documentation lists is a metric for products saved to a cart.

So a shopper could park your product in Universal Cart for three weeks, waiting patiently for you to blink on price, and on current evidence you won't see that intent until it converts or quietly evaporates. Google will know. You'll find out at the till, or never.

Shopify merchants get a bonus round. Shopify's docs for its Google AI Mode and Gemini channel say Direct Checkout is on by default for eligible US stores, and Google Analytics and custom pixels don't fire on those purchases. Nobody has confirmed Universal Cart follows the same flow, so let's not assume. But if you haven't looked at the Agentic settings under sales channels in your Shopify admin, this would be a lovely week to start. (Default settings: the silent co-founders of every ecommerce business.)

A product page isn't just a till. It's where you capture the email, offer the bundle, tell the brand story and enrol a first-time buyer into the flows that make order two cheaper than order one. Native checkout inside Google bypasses most of that. Your lovingly art-directed product page shrinks to a set of attributes, rendered in Google's design language, right next to your competitors' attributes.

Amazon sellers are nodding wearily by now, because that's the marketplace bargain. On Amazon, the third-party seller is technically the seller. You own the inventory and the risk. Amazon owns the page template, anonymises the buyer, restricts off-platform marketing and keeps the behavioural data powering its modestly sized advertising business. Sellers sign up anyway, because the demand is enormous.

The question was never are you the seller of record? It was what did you trade for the traffic?

Universal Cart applies a similar trade to the open web, with one difference. Brands entered Amazon's bargain knowingly. Some merchants will enter Google's via a default toggle, or through a product feed they already maintain for Shopping ads. The feed built to win clicks is quietly becoming the storefront.

To be fair to Google (it happens), cart transfer exists precisely so merchants can keep checkout, upsells and consent capture on their own site. And Google has every reason not to annoy the retailers who populate its Shopping Graph.

The framing has also shifted. At I/O in May, Universal Cart was pitched as the piece tying Google's agentic commerce work together. By October, Google had neatly separated features that act on a shopper's behalf from tools that give shoppers "confidence and control", and filed the cart under the second. Price history and loyalty information, both mentioned in May, didn't make the launch post. Could be a phased rollout. Could be Google stepping carefully around how much autonomy it admits to. (Nothing says "we're completely relaxed about this" like a quiet reclassification five months in.)

This is the change I'd circle in red pen, and it's barely been discussed.

For most DTC brands, abandoned cart recovery is among their most profitable flows. Shopper wanders off, your email nudges them back, often without any discount.

Universal Cart relocates the abandoned cart. A shopper who saves items inside Google and walks away has abandoned Google's cart, not yours. What brings them back isn't the reminder email your copywriter agonised over. It's Google's notification. And the triggers Google has described are a lower price, a restock, or an available promo code.

Translation: the default recovery lever inside Universal Cart is a margin concession. A sale you'd once have rescued with a well-timed email might now be rescued, if at all, by cutting your price and letting Google helpfully broadcast it to every shopper holding your product. Your retention flow has been replaced by a discount siren.

Google's timing is, let's say, efficient. Alongside the cart came its first "When to Buy" guide, mapping when discounts historically peak across US categories.

Some findings are unsurprising. Footwear, TVs and audio peak over Black Friday, which Google defines as the Tuesday before Thanksgiving through Cyber Monday (24 to 30 November this year). Others are more pointed. Handbags, kitchenware, large luggage and electronic toys hit their deepest markdowns in late December, and the guide advises deal-focused handbag shoppers to skip Black Friday altogether.

Read that from the merchant's side. Google is teaching shoppers to wait in specific categories, at the same moment it hands them a cart that will tell them exactly when to come back. A markdown calendar that used to live in the heads of retail buyers has been published to consumers by the world's largest search engine. To be scrupulously fair, the guide calls itself guidance, not a guarantee.

Same week, Google also gave links more prominence in AI Mode and AI Overviews. Google says Preferred Sources usage has grown more than sevenfold since January, without publishing a baseline or explaining the measurement. (Sevenfold of what, exactly, remains a private matter between Google and its spreadsheet.)

A tidy division of labour appears. Informational queries, where publishers shouted loudest, get more visible links. Commercial queries get infrastructure that lets shoppers finish without leaving. Your content might earn more citations in AI answers this quarter while your product pages get fewer visits.

One caveat: Search Console gives every link in an AI Overview the overview's single position, so climbing within it won't show. Do warn whoever builds the monthly deck.

As promised, Amazon. Its absence tracks with its behaviour since mid-2025: it pulled out of Google Shopping ads, blocked AI shopping agents from Google, OpenAI, Perplexity and others, and built its own agentic tools in Rufus and Buy for Me. Some reporting suggests it has stopped supplying Shopping feeds too, on thin sourcing.

For Amazon-only sellers, Universal Cart is a channel your listings can't reach, which, given everything above, isn't necessarily a problem yet. But if shoppers adopt a cross-retailer cart sitting upstream of every store, brands with a Shopify store or a presence on Walmart or Target gain a route into those baskets that pure Amazon sellers lack. If shoppers shrug, Amazon's decision to stay out will look rather clever.

None of this argues for opting out on reflex; that may just hand the visibility to a competitor with a cleaner feed. The more defensible stance is to treat participation as a set of deliberate choices rather than a default:

Open questions remain. Will merchants ever see the saved-but-unpurchased products sitting in Google's carts? Will the YouTube and Gmail integrations promised in May actually turn up? Will one-way transfer become two-way sync, and on whose terms? And if shoppers come to expect one basket across every store, does staying out stop being a choice and start being a cost?

Amazon sellers have spent a decade learning the difference between owning a sale and owning a customer. This holiday season, a much larger group of brands is about to sit the same exam.

Jo

P.S. "Seller of record" is the ecommerce equivalent of being called "a valued member of the team" in the same meeting where they announce the restructure.

P.P.S. If anyone finds a saved-to-cart metric anywhere in Merchant Center, reply immediately. I'll have it framed.

📝 Copy AI - Generate content, such as blog headlines, emails, and social media posts using AI

📊 Supadash - Connect your database and instantly get AI-generated charts to visualize your data and track metrics about your application.

🔄 BypassGPT - A tool to rewrite AI-generated content for SEO and undetectable by AI detectors.

🖥️ Tiledesk - Rapidly deploy AI-powered conversations across social channels.

😍It was great, all I needed

🙂Good, not great

🙄It sucked