Google’s Agentic Wallet: Is the Fifth Time the Charm?

That brings me to Google and its announcement last week at Google I/O about AI and the revival of Google’s Wallet ambitions. Now on its fifth installment.

Google Wallet, the original, launched in 2011. It flopped. The reboot, Android Pay, arrived in 2015. That morphed into Google Pay in 2018. Google Pay, the wallet, was upgraded in 2020, complete with bank accounts and peer-to-peer transfers, then was scaled back and shut down in the U.S. in 2024. The IP was handed back to a revived Google Wallet. In between, Google Shopping, Google Express and Google Actions each tried to build a commerce storyline and were written out of the script.

Five tries. Same storyline every time. Google’s scale plus payments equals commerce on steroids. A slam dunk. Same disappointing ending with every sequel. The advertising business, which actually rings the cash register, swallows the commerce ambition whole.

At Google I/O 2026, Google Wallet with Google Pay plays a starring role in furthering the hottest commerce storyline in thirty years. Agentic commerce. And the reviews are in.

Following Google’s announcements that day, the vast majority of the 63 analysts who cover Alphabet rated it a buy. Mizuho’s Lloyd Walmsley wrote that the company has shifted “from AI Loser to AI Winner and deserves a premium” and set his price target at 30x his 2027 GAAP EPS estimate.

The critics love the fifth installment. But the audience sitting in the seats has seen this movie, and its four previous sequels, before.

The Google AI Wallet Plot

The production values, it must be said, have the makings of a blockbuster. Gemini 3.5 Flash is now the default model across the Gemini app and Search’s AI Mode. Sundar Pichai said that the Gemini app serves more than 900 million monthly users across 230 countries. Daily queries, he claims, have grown sevenfold. Google is processing more than three trillion tokens a day internally for its own developers.

But the plot point that matters to commerce is what Google did with Google Wallet.

Google introduced Universal Cart, an AI-powered shopping hub that lets users add products from Search, Gemini, YouTube and Gmail into a single cart. A Gemini-powered agent embedded inside the cart monitors price drops, surfaces price history, sends back-in-stock alerts and runs compatibility checks across items.

Because Universal Cart is built on Google Wallet, that agent has access to the consumer’s stored payment credentials, loyalty information and merchant offers, and uses that intelligence to recommend the best way to pay. Google Pay, the transaction layer inside Wallet, completes the purchase through one-tap payment or transfer to the retailer’s own site. Gemini is the brain. Google Wallet is the credential vault. Google Pay executes the purchase. All of it aimed at positioning Google at the front end of the agentic transaction.

Read More: The Rise of AI Agents and the Fall of Digital Wallets

Google’s Universal Commerce Protocol (UCP), co-developed with Shopify, creates an open standard for agentic checkout. The Agent Payments Protocol (AP2), built with more than 60 partners including PayPal, Mastercard and American Express, lets AI agents make purchases on the consumer’s behalf using cryptographically signed digital contracts. Launch partners include Nike, Sephora, Target, Ulta Beauty, Walmart and Wayfair.

In other words, Google has assembled the pieces of a personal commerce assistant, something that has eluded them: a discovery engine, a credential vault, an agent brain and a payments rail. The question is whether assembling pieces is the same thing as building a commerce platform.

It’s worth reminding readers that Google has had most of these pieces before. It had the search data. It had the payment credentials. It had the merchant relationships. It even had loyalty programs and offers baked into Google Pay.

And it still couldn’t turn any of it into a commerce platform that consumers or merchants wrapped their digital arms around. The advertising business ate every commerce initiative alive because Google’s incentive to monetize intent through ad placement always overwhelmed the incentive to close the sale on behalf of the consumer.

What’s Different This Time. And What’s Not

Every previous Google Wallet play was tethered to Android. Android’s roughly 100 million U.S. users skew lower-income and lower-spending than Apple’s 155 million iPhone users. Analyst Horace Dediu calculates that iPhone users spend about $11.20 per month on apps versus $1.70 for Android users: 7.4 times more valuable on a per-user basis. Building a commerce empire on the platform with fewer users spending less money is the equivalent of being at the starting line of the Indy 500 with three flat tires.

This time, the Agentic Google Wallet runs through Chrome. And that, they hope, may change the math.

Google is merging Chrome autofill into Google Wallet, collapsing the distinction between “card saved in my browser” and “credential in my wallet.” Chrome already shows card benefits, loyalty card numbers and travel details alongside saved payment methods, and that integration is now expanding to desktop and iOS.

At Google I/O 2026, Google went further, expanding Chrome’s autofill to pull passports, driver’s licenses, boarding passes and booking confirmations directly from Google Wallet across every platform. It also introduced Cross-device Payment Verification, letting a desktop Chrome user approve a web purchase by tapping their Android phone with biometric confirmation. Those aren’t wallet features. They’re credential infrastructure features, embedded inside a browser 3 billion people already use, and that most of them don’t think of as a payments product at all.

Read More: The Battle for AI Isn’t About Models. It’s About Habits

You also don’t have to be a Gemini app user to enter Google’s agentic commerce funnel.

Google is rolling Universal Cart out in Search AI Mode, with YouTube and Gmail to follow. Gemini is already built into Chrome on iOS. That means an iPhone user who searches on Google in Chrome, watches YouTube and checks Gmail is inside the funnel without ever downloading the Gemini app. When that user checks out, the credentials come from Google Wallet via Chrome. Not Apple Pay.

That’s the theory, anyway.

Except Apple blew a hole in it. With iOS 26, Apple expanded its Wallet AutoFill beyond Safari to work system-wide, including inside Chrome. When an iPhone user fills out a payment form in Chrome, Apple’s Wallet now competes directly with Google’s autofill to surface the credential. And Apple’s version is tightly integrated with Face ID, pulling card details from a Wallet app that iPhone users already trust, already use and already have set up. The Chrome-as-Trojan-horse thesis depended on Google’s autofill being the default path to the credential inside Chrome on the iPhone. Apple’s upgrade puts Apple’s own Wallet in front of that path, intercepting the consumer at exactly the moment Google was counting on capturing them.

Chrome holds roughly 50% of the U.S. browser market and more than 62% of desktop traffic. But on the device that matters most for commerce, the iPhone, Apple now owns the autofill prompt in every app, including Chrome. The landscape has shifted from who owns the NFC tap to who owns the autofill prompt. And in that version of the sequel, Apple may have just moved to the front of the line.

But the Chrome problem is a sideshow compared to the bigger foundational issue nobody is stress-testing.

All of this assumes people will keep using Google to search. And our PYMNTS Intelligence data says that assumption may be flawed.

The Search Exodus

PYMNTS Intelligence research from April 2026 found that 38% of consumers have already used dedicated AI platforms for product discovery. More striking, more than a third (35%) say they have fully replaced their prior methods. They’re not layering AI on top of old habits. They shut the door on the old way and left.

And they’re not coming back.

Read More: Why 30 Million US Consumers No Longer Search

Among those who have replaced their old methods with generative AI, PYMNTS Intelligence reports 39%  using search engines less. More than a third (36%) visit retail sites less often. Nearly a third of replacers, 30%, say AI platforms are the most helpful tool to guide their purchases. These are not people toggling between Google and ChatGPT. These are people who tried something new and decided it was better.

About 9% of U.S. consumers, roughly 24 million people, are what our PYMNTS Intelligence research calls Prompt Economy Pros. These power users have replaced their old routines with AI across shopping, finance, health, travel and learning. They perform an average of twenty-five different tasks with generative AI. Among AI power users, 54% have replaced their old methods of product discovery. Here we see Platform Dynamics 101 in action. The more someone uses a platform, the harder it becomes to leave, because the platform learns their preferences, holds their history, and fits into their daily routine.

Universal Cart depends on Google Search as its primary on ramp. If discovery keeps migrating away from Google, the cart will have no one to fill it.

And the migration has a compounding problem that predates agentic anything.

Amazon.

Depending on your source, anywhere from a third to more than half of U.S. product searches already start on Amazon, not Google. And Amazon’s advantage isn’t just in discovery. It’s end-to-end. Discovery, purchase, fulfillment, returns and customer service, all inside one ecosystem. Amazon controls the entire journey from idea and intent to arrival in the home. Google controls none of that. Google can help a consumer find a product. It can’t sell it to you. It can’t ship it. It can’t take it back when it doesn’t fit.

Amazon’s real strength, as I’ve written at PYMNTS, lies in its end-to-end control of the commerce stack. Google is still, at its core, an aggregation layer. A very sophisticated one, now with AI and a wallet attached. But an aggregation layer all the same.

That distinction matters enormously in an agentic world. Because when something goes wrong, someone has to own the problem. And that is the big gap that no protocol or partnership announcement can close.

The Accountability Gap

So here’s a question that got zero airtime at Google I/O. If Gemini recommends a product, helps a consumer add it to Universal Cart, selects the payment credential and completes the checkout through Google Pay, and then the product arrives damaged, or late, or never arrives at all, who does the consumer call?

Not Google. Google isn’t the merchant of record. Google didn’t fulfill the order. Google didn’t process the return. Google’s agent made the recommendation, chose the payment method and executed the transaction. But when the experience breaks, the consumer is on her own, navigating the merchant’s customer service, the issuer’s dispute process and whatever return policy applies, none of which the Gemini agent controls.

Read More: Department Stores of the Future Are AI Agents

Without end-to-end control of intent, purchase and fulfillment, the agent stands to take the blame for things that go wrong while having zero ability to fix them. That’s a trust problem that gets worse, not better, as the agent takes on more autonomy. The more consumers let Gemini act on their behalf, the more they will hold Gemini responsible for the outcomes. And Google has no mechanism to make those outcomes right.

Amazon’s agent doesn’t have this problem. When Alexa helps a consumer buy something on Amazon and the product arrives wrong, the consumer calls Amazon. Amazon controls the fulfillment. Amazon processes the return. Amazon issues the refund. The agent and the commerce platform are the same entity. That is the structural advantage of end-to-end control, and it is exactly the advantage Google does not have.

Monetizing Intent

Every Google commerce initiative that’s come before has suffered the same demise. Not because the technology failed. Because the advertising business ate its lunch.

Think about what the Gemini agent inside Universal Cart now captures and why Google is so excited about it. What the consumer searched for, what they compared, what they rejected, what YouTube video prompted the purchase, what Gmail promotion caught their attention, what price they were willing to pay, what offer made them convert. The complete map of consumer shopping intent, the stuff inside their head, from first impulse to final click.

Google’s always made its money by capturing data about what people want and selling access to that data to advertisers using Search. Now Google is adding the most valuable data of all, the actual purchase decision, connected to the discovery prompts that led the consumer there.

The result is a closed loop of commercial intent that no other company in the world possesses. At scale.

The merchant of record still processes the order. The issuer’s card still settles the payment. Google says merchants keep their customer data. Technically true. But neither the merchant nor the issuer sees the intent data that preceded the transaction. Google sees it all.

A merchant who joins Universal Cart gets access to Google’s consumers. Google gets access to the merchant’s customers’ complete decision journey. And then Google monetizes that journey, the searches, the comparisons, the abandoned carts, the price sensitivity signals, by selling ads to that merchant’s competitors.

Even embedded inside the prompt.

Merchants have watched this movie. All four previous installments. They spent two decades paying Google to reach their own customers through search ads. Now Google is offering to sit inside the transaction itself. And merchants are supposed to be okay with access to transaction data, the richest data set around, that could be used against them.

Here’s the real-life test. When a consumer asks Gemini to find the best running shoe under $150, whose interest does the agent serve? The consumer’s, finding the objectively best product at the best price? Or the merchant’s, the one who paid Google for preferential placement inside AI Mode?

And that comes down to who the consumer trusts to act on their behalf.

That’s still fundamentally the same Google. An advertising engine that monetizes intent via search. The packaging is different. The AI is better. But the business model hasn’t changed. And every time Google has had the chance to put the consumer’s commerce interest ahead of the advertiser’s revenue interest, the advertiser has won.

Every single time. Through four previous sequels.

What Consumers Want and What They Don’t

The consumer data makes the trust question inescapable.

Our PYMNTS Intelligence research shows that nearly 70% of consumers are interested in using AI agents for shopping-related tasks. But not all agent delegation is created equal.

Read More: 70% of Consumers Say Yes to AI Agents for Shopping

Consumers want smarter credentials and smarter discovery. They want AI to search for products, compare options, track prices, manage loyalty programs, find the best payment terms and surface deals they would have missed. Sixty-nine percent will hand off subscription management. They want an agent that functions like a good personal assistant, one that knows their preferences, remembers their history and optimizes their money.

What they don’t want is to hand over the keys. Not yet. And they especially don’t want an agent that’s working for someone else out of their line of sight.

Our PYMNTS Intelligence data from the 2026 Global Digital Shopping Index, conducted in collaboration with Visa Acceptance Solutions and based on surveys of more than 5,200 consumers, 1,185 merchants, and 150 acquirers across three countries, makes this unmistakable.

Only 6% want full automation of the payment decision. At a $300 purchase, 52% want no AI involvement at all in selecting how they pay. Only 22% would let an agent auto-order even with a 24-hour cancellation window. Thirty-nine percent reject autonomous purchasing under any conditions.

Consumers want AI to optimize, not to act. They want the agent to do the homework and present the best answer, and then they want to make the final call.

That’s a real opening for anyone who can build a credential layer that genuinely serves the consumer’s interest.

It’s a big problem for anyone whose business model is built on serving the advertiser’s.

The Smart Credential Is the New Wallet

This is the thesis I keep coming back to, the one I laid out nine months ago as V.1 of agentic shopping was getting a head of steam. That agentic commerce success is anchored in who owns the credential, the intermediary that completes the transaction, the interface that gives the model permission to spend. The operating rules that protect the consumer in case a size 42 showed up instead of a size 24.

An AI model without a trusted and secure credential is a shopper without a credit card. It can window-shop. It can’t buy.

Read More: Why AI Shopping Is Still Just a Smarter Search Bar

The entire agentic commerce architecture — Google’s UCP, its AP2, Mastercard’s Agent Pay, OpenAI’s Agentic Commerce Protocol, Visa’s Intelligent Commerce — converges on the same fundamental truth. The credential is the critical chokepoint. And Google just made a massive move to own that critical chokepoint across platforms and operating systems.

But here’s what makes the credential story more nuanced than “whoever owns the wallet wins.”

The wallet as consumers know it, the button, the screen, the “pay with” prompt, disappears in an agentic world. When an AI agent calls a credential programmatically and the transaction completes without the consumer ever seeing a payment interface, the wallet doesn’t die. It changes from a consumer-facing checkout button to a machine-facing authorization layer.

Agentic commerce doesn’t need a wallet button. It needs a credential that an agent can call.

And that credential has to be smart.

Read More: Why OpenAI, Amazon and Apple Want to Be the Smartphone in Your Pocket

As I have written previously, long before anyone in the industry started talking about agentic commerce, consumers had already built their own manual version of a smart credential. They use a revolving credit line to absorb unexpected expenses and bridge pay cycles. They convert a big-ticket purchase into card installments. They route a smaller purchase through BNPL for predictable repayment. A grocery purchase late in the pay cycle goes on one account to preserve cash. A furniture purchase gets routed to installments. A merchant promotion redirects to embedded financing.

What looks like random acts of credit usage to the industry is actually a personal credit stack that consumers manage manually, with incomplete information, under time pressure.

Read More: The Next Battle in Credit Won’t Be for Top of Wallet

The smart credential automates that stack and removes the friction.

And the credential that carries enough intelligence to optimize across the consumer’s full financial picture, not just which card to charge, but which payment method, which terms, which merchant offers to apply, is the one the agent will prefer because it is the one that its user will continue to favor.

Top of wallet becomes top of agent. And top of agent is decided by the machine, not the human.

This is where card networks and issuers have an opening that neither Google nor Apple can match.

The networks don’t sell advertising. They don’t compete with merchants for consumer attention. They can align all three sides of the transaction, the issuer, the merchant, and the consumer, without the advertising conflict that makes Google’s model fundamentally challenging as a trust layer.

Both Mastercard and Visa are already inside Google’s protocols. And both are building the infrastructure that lets issuers turn static credentials into dynamic smart ones. Flexible credentials that simplify payment options into a single account with multiple ways to settle, carrying merchant offers, identity, loyalty intelligence and optimized payment routing directly to the agent.

When a consumer’s agent shops on her behalf, the issuer’s credential can abstract the complexity within a few milliseconds. “This merchant has a 10% offer for this cardholder. This purchase qualifies for 3x points. This transaction should route to the installment plan.”

The credential carries the intelligence. The agent has a reason to prefer it. And the consumer gets a better financial outcome without doing the work.

Read More: Why the Offers Economy Is Broken

Merchants have an incentive to participate since their offers are in the purchase flow. Those offers can be funded by brands who want more attribution for sales of their products to customers. Merchant advertising dollars can be redirected and funded by brands and issuers that deliver higher conversion.

The issuers whose credentials are smart enough to participate in that highly optimized feedback loop will capture disproportionate volume. The ones whose credentials are little more than dumb pipes will become the Bing of payments. Technically functional, rarely chosen.

Google has the reach. The networks and issuers have neutrality and the three-sided alignment. The question is whether an ad-supported company is the right steward of a credential that’s supposed to work for the consumer.

Roll the Agentic Credits

In Hollywood, when a franchise keeps getting re-upped, it usually means the studio believes in the idea but has never found the right story. Google has been trying to own digital payments since 2011 through Google Wallet, Android Pay, Google Pay and more. The core idea, becoming the thing you tap or click every time you buy something, is a great idea.

This latest version, built into Chrome and powered by AI agents that help consumers shop, gives Google a reach and a merchant pitch it’s never had before.

But Google’s playing catch-up. And this time it’s playing catch-up from a weaker position than many of the analysts seem to appreciate.

ChatGPT is the AI that most people actually use, with 900 million weekly active users globally as of early 2026, up from 500 million a year earlier. Our PYMNTS Intelligence finds that more than 7 in ten U.S. AI users have tried it at least once, and that number has held steadily high since December 2025. Google’s Gemini comes in second at just over half that adoption, and ChatGPT is nearly three times larger on both web and mobile.

More interesting is the behavior pattern underneath.

Read More: The Battle for AI Isn’t About Models. It’s About Habits

People pick a go-to AI the same way they pick a go-to store. And when they do, it’s usually ChatGPT, the way it’s usually Amazon, and then they occasionally try others for specific needs. The go-to gets most of the attention, most of the repeat visits and most of the trust.

What matters about being the default isn’t just that people show up. It’s whether they stay. Whether the default becomes the habit. Whether the habit becomes the place they do everything.

That’s already happening in AI. The more someone uses a platform, the harder it becomes to leave because the platform learns their preferences, holds their history and fits into their daily routine.

We’ve seen this before. It’s how mobile apps replaced desktop websites.

This is the world Google’s trying to break into. And what it’s building does match what consumers say they want. Make shopping smarter, not create a robot that buys things for them.

But look at what Google’s actually built. A system that connects what consumers search for, what they decide to buy and how they pay for it. All of it running through Google and embedded into its operating system. In this setup, neither the store nor the card issuer nor the network on which those credentials run sees the most valuable piece. The context of why that consumer bought what they bought.

Read More: 65% Call Insurance Essential. Why Most Spending Isn’t So Clear-Cut 

That’s not a shopping tool. That’s an advertising engine with a pay button pushed by an agent.

Google’s foundational problems have only gotten harder since the last sequel.

Consumers are leaving Google Search for AI platforms and not coming back. Amazon dominates product search and owns the full stack from discovery to doorstep. Apple’s iOS 26 upgrade muscles its way in front of Chrome for payment autofill on the iPhone. The agent has no mechanism to fix a broken order. And every previous time Google has had this kind of data, it turned it into ad targeting, not consumer commerce.

The companies that win in AI-powered shopping will be the ones that serve the payments ecosystem without a hidden advertising motive. That earn the trust of consumers who have made it clear they want an AI that works for them, not for an ad platform.

Five tries in, Google hasn’t proven it can be that company. Not yet. This time, though, it certainly has a better script. Now, we’ll sit back to see whether the audience watches the sequel until the very end.

Until NEXT time.

Join the 21,000 subscribers who’ve already said yes to what’s NEXT.