Google Just Made Your Loyalty Programme Portable.
That's Not the Good News It Sounds Like.
Mark Sage - 5 min read - 27/08/2026
There's a detail in Google's Universal Commerce Protocol that deserves more attention than it's getting.
UCP — the open standard Google has built with Shopify, Walmart, Target and others to let AI agents shop on your behalf — recently gained what Google calls identity linking. In plain terms: when your AI agent buys something for you inside Gemini, your loyalty benefits travel with it. Member pricing, points, free shipping — recognised and applied automatically, inside Google's surface, without you ever visiting the retailer's own site or app.
Google frames this as a win for brands: it protects the value of the direct customer relationship. I'd read it the other way around. Loyalty programmes were the last structural reason for a customer to come to you directly. If points and perks work just as well inside someone else's agent, that reason quietly evaporates. The drawbridge has been lowered from the inside.
To see where this leads, it's worth going somewhere this future has already happened.
The Table That Belongs to Someone Else
Arriving in China as a non-resident, the ecosystem feels less like convenience and more like a locked door — platforms, payments, merchants, all interconnected in ways your instincts don't expect. Getting set up is daunting. Then, once you're in, it's remarkably smooth.
F&B is the best place to feel it. You sit down at a restaurant table and there's a QR code. Nothing unusual in that — since COVID, QR codes on tables are common worldwide. What's different in China is that the code is typically platform-dependent. You scan it with Meituan or Alipay, and from there you browse the menu, order, and pay — natively integrated or seamlessly linked. The ticket is raised in the kitchen. The food arrives. No menu, no till, no interaction.
The interesting part is the channel integration. That same Meituan app handles your delivery at home, your pick-up on the way past, and your table service when you dine in. One ecosystem sees everything the customer orders, at every restaurant, across every occasion.
The customer may be eating at the restaurant. But they belong to Meituan.
It's tempting to treat this as something new — a distinctly Chinese model. The reality is that the West has had a similarly connected network, hoovering up customers, for decades.
It's Visa and Mastercard.
You can buy almost nothing, online or off, without touching one of these networks. That gives them vast data visibility and, more importantly, network power over merchants: opt out of the network and you effectively opt out of the market.
There's a lesson in how that story unfolded. The networks began by providing genuine benefits — reduced friction, customer access, easier operations — and merchants joined willingly. Then the power shifted, and the networks came to own the traffic and the tax on it. Regulators in the EU and Australia have spent years — decades — slowly, very slowly, forcing that tax down. The tax gets capped. The position never does.
Other markets show what happens when the rails themselves become public infrastructure: India's UPI and Brazil's Pix run at effectively zero tax. But the network doesn't disappear — it moves up the stack, to Google Pay or PhonePe. Commoditise the rail, and the chokepoint simply relocates to whoever owns the interface.
The Difference is in the Data
China, though, shows what happens when a network gains not just scale but data. The card networks see the merchant, the amount, and the time. They never saw the basket. Platforms like Meituan see the identified customer — and exactly what they buy, where, when, and how often, across every channel and every merchant. That's the difference between knowing a transaction happened and knowing a person.
That level of visibility is now arriving in the West. Which brings us back to UCP.
Google is careful to say that merchants "remain the Merchant of Record" — your commerce, your customer. But this conflates the legal customer relationship with the behavioural one. If discovery, comparison, cart, and checkout all happen inside Gemini, the merchant owns a record of a customer they no longer actually meet. The cross-merchant intelligence — that Mark compared three options, is price-sensitive on Tuesdays, responds to free-shipping nudges — accrues to the surface owner. Universal Cart literally aggregates shopping across retailers into a single Google-owned layer. And now, with identity linking, even your loyalty identity — the one thing that was supposed to bind the customer to you — authenticates through their front door.
Then comes the monetisation. The protocol is free and open today; the advertising integrations have already begun. Search was free too — the toll booth gets built after the road becomes indispensable. The expansion map reads like the Meituan playbook: food delivery, hotel bookings, and a rollout targeting markets including Australia and the UK. And the most telling detail of all? Among UCP's endorsers are Visa and Mastercard. The oldest networks in the puzzle, making sure they're still inside the flow when the new middleman forms. They, of all players, know how this game ends.
It's worth asking why Google is doing this at all. Google is one of today's aggregators — and agentic commerce is an existential question for it. If an AI agent mediates buying, what happens to search ads, to Shopping, to the attention economy Google taxes? UCP isn't just an attempt to invent agentic commerce. It's an attempt to remain relevant in commerce at all.
The Chinese platforms see the same threat from the other side. If my agent can order directly from any merchant, why route through Meituan's app and pay its commission? Meituan is as exposed to agentic disintermediation as the noodle shop was to Meituan — which is why every super-app is racing to be the agentic layer. Each interface generation re-runs the aggregation game, and the chokepoint moves: from payment, to search, to the app, to intent.
The Luckin Counter Play
Which brings us back to Luckin Coffee, which I wrote about previously.
Their play has been the opposite: own the network. Customers order direct, on Luckin's rails, inside Luckin's platform — no intermediary owns that relationship. Customers make that effort because the product is worth it, just as customers pay more for Starbucks because the "third place" means something to them. Brand, product, and experience are what make owning your platform viable.
But there's a caveat that should focus minds: this only works while the human is doing the choosing. If I simply tell my agent "I want a coconut latte," the agent decides where it comes from. Agents don't feel warmth. They don't value the third place. They optimise on price, speed, rating, and availability — and whoever is most legible and most competitive to the agent wins the order. Every layer of brand equity built for human choosers gets discounted the moment the chooser isn't human.
So the strategic question for any merchant isn't whether to engage with these networks — it's how much, and how far. How much of your customer are you willing to accept becoming theirs?
Building your own platform, as Luckin did, is one answer. Building your own agentic presence, as Woolworths is doing, is another. Both are bets on the same thing: staying relevant, and staying connected to your customer, when the interface between you and them is increasingly owned by someone else.
And as for that loyalty integration — when your points work inside someone else's agent, it's worth asking whose loyalty programme it really is.
In China, the QR code on the table looks like convenience. It is convenience. It's also a transfer. With agentic commerce and UCP, that convenience is arriving here — loyalty points and all. It is also a transfer. And after forty years of card networks, we of all people should know how that story goes.