Yesterday, Synchrony Financial announced an enterprise collaboration with OpenAI “to strengthen Synchrony’s positioning at the center of AI’s next chapter in shopping and payments.” The press release is dense with corporate strategy language, but the operative sentence is this one: the collaboration exists to bring “financing, rewards, and loyalty into AI-native shopping and checkout experiences.”

Buried in the announcement is the part that matters. Synchrony has launched a plugin in the ChatGPT plugin directory that lets consumers “discover savings and offers within the Synchrony Marketplace directly within ChatGPT and browse promotional financing, deals and everyday value from participating Synchrony partners in a fast, conversational experience.”

Read that again. The lender is now inside the shopping conversation. Not at checkout. Not in a separate app. In the same interface where an AI agent is recommending, comparing, and selecting products.

The agentic commerce stack now has a standardized payment layer, an embedded financing layer, and an incumbent loyalty layer. The only layer nobody has shipped is the one that verifies whether the product being financed, purchased, and rewarded is actually good. That gap is no longer a philosophical objection. It is the load-bearing wall everyone forgot to pour.

What Synchrony Actually Is

If you do not follow consumer finance, the announcement reads like a mid-size bank doing an AI partnership. It is not. Synchrony is the private-label credit infrastructure of American retail. Its co-brand and store card partners include Amazon, Lowe’s, Verizon, Walgreens, Sam’s Club, and Cathay Pacific, alongside hundreds of thousands of smaller merchant locations. It holds roughly $119.5 billion in assets, generated $24.17 billion in revenue in 2024, and has served as the exclusive issuer of PayPal Credit in the United States since acquiring that $7.6 billion receivables portfolio in 2018, an arrangement contractually locked in through 2028.

When Synchrony says it sits at the intersection of “consumer financing, payments, loyalty, and merchant partnerships,” that is not marketing. It is a description of the credit plumbing behind a large fraction of everyday US retail. Tens of millions of consumers carry its paper, most of them without ever having heard the company’s name.

Which is precisely why this partnership is significant. Synchrony does not need to win the agent wars. It needs to make sure that whatever agent wins, its financing is attached to the transaction. The ChatGPT plugin is that attachment point.

The Checkout Wars Are Over. The Protocols Won.

To understand why a lender is moving now, look at the timeline Harvard Business Review laid out in its August 7 piece on algorithmic shopping:

  • April 2025: Amazon launches “Buy for Me,” letting its AI agent visit brand websites, select products, enter payment details, and complete purchases without the customer leaving the Amazon app.
  • September 2025: OpenAI introduces Instant Checkout inside ChatGPT, built on the open-source Agentic Commerce Protocol.
  • January 2026: Google unveils the Universal Commerce Protocol at the National Retail Federation conference, an open standard built with Shopify, Target, Walmart, and more than 20 other partners, covering the process “from discovery to post-purchase support.”
  • January 2026: Microsoft launches Copilot Checkout at the same conference, and Shopify switches on agent-readable storefronts by default across millions of merchants.

In eleven months, the industry standardized how an AI agent pays. Gartner now projects that by 2028, 90% of B2B purchases, more than $15 trillion, will flow through AI agent exchanges.

Every one of those protocols solves the last step of the transaction: authentication, payment instrument, receipt, returns. None of them addresses the step before it: how the agent decides what to buy. The protocols treat product selection as a solved input. It is the least solved input in the entire stack.

Why the Financing Layer Is Different From the Payment Layer

A payment protocol is neutral infrastructure. Stripe, the ACP, the Universal Commerce Protocol: they move money and move it back on refunds. They do not care what you buy.

Financing is not neutral. Financing is the most powerful conversion lever in retail, and it has been for decades. Deferred-interest promotions raise average order values on big-ticket purchases. “Six months, no interest” turns a $900 deliberation into a $2,400 checkout. Buy-now-pay-later removes the price signal at the exact moment of decision. Retailers pay lenders interchange-like economics precisely because financing converts browsing into spending.

Now that machinery has an API into the conversation where products get chosen. The Synchrony plugin does not wait for checkout. It surfaces “promotional financing, deals and everyday value” while the user is still asking ChatGPT what to buy. The offer becomes part of the consideration set. The agent that recommends the product and the financing that pays for it now live in the same window, and one of them earns money when the purchase happens.

OpenAI’s own framing makes the ambition explicit. Kaylin Voss, VP of Americas and Industries at OpenAI, described the opportunity as “reimagin[ing] the entire commerce experience: from how customers discover products to how they pay, earn rewards, and build loyalty.” Maran Nalluswami, Synchrony’s EVP and Chief Strategy and Business Development Officer, said the collaboration is meant to ensure “the value they’ve entrusted in Synchrony products will thrive in the agentic commerce era.”

Discover, pay, earn rewards. Notice what is not in that sentence: verify.

The Incentive Problem, Stated Plainly

Sponsored listings taught us what happens when the ranking layer is paid for. Amazon search results blend organic placement with sponsored slots, and an entire optimization industry grew up to game the blend. Star ratings taught us what happens when the trust signal is gamifiable: the FTC now has a rule banning fake reviews outright, and it has spent the past year in enforcement actions against the infrastructure that sells them.

Agent-mediated financing is the third iteration of the same pattern, with a sharper edge. A sponsored listing can only bias which product you see. A financing offer can change what you can afford to buy. It expands the budget constraint inside the conversation, before any human deliberation happens. And unlike a banner ad, a financing nudge arrives framed as value: “you have a promotional offer available,” delivered by an assistant the consumer now trusts more than newspapers, as the RTB House survey we covered last week documented, with 44% of US consumers trusting AI tools for purchase decisions.

Put the two findings side by side:

  • Consumers trust AI shopping tools more than every media channel except friends and family.
  • The companies that profit from purchase volume are now plugging offers directly into those tools.

This is not an accusation against Synchrony or OpenAI. Promotional financing is legal, disclosed, and genuinely useful when used responsibly. It is an observation about structure. When the most trusted channel in commerce becomes a distribution surface for the most potent conversion lever in retail, the integrity of the channel’s product information stops being a nice-to-have. It becomes the only counterweight in the system.

The Regulator Asymmetry

There is also a governance gap, and it has a shape.

Product-side deception is being policed aggressively. The FTC’s fake review rule carries civil penalties. Its 2026 enforcement docket has targeted review brokers, deceptive AI claims, and scalper bots. The product recommendation layer, at least, has a sheriff.

The agent-mediated credit layer has none, and its incoming participants carry relevant histories. Synchrony’s corporate lineage includes a $225 million settlement with the Consumer Financial Protection Bureau and the Department of Justice in 2014 over deceptive and discriminatory credit card practices at its GE Capital predecessor. Consumer advocacy groups have pressed the CFPB over PayPal Credit’s use for for-profit school tuition, classifying its 24% interest rates and collection practices as predatory. A 2024 class action alleged CareCredit’s 32.99% APR violates New York usury laws. None of this makes Synchrony a bad actor today. All of it is a reminder that consumer credit is the most regulated consumer product in America for a reason, and that reason is that the industry’s incentive to expand borrowing reliably outruns consumers’ ability to price it.

Now transpose that dynamic into an interface with no APR table, no Schumer box, no checkout page where the terms sit in a legally mandated rectangle. A conversational financing offer, delivered mid-recommendation by a trusted assistant, is exactly the environment disclosure rules were never designed for. The CFPB has started probing AI-driven credit marketing, but no framework exists for “agent recommends product plus financing package” as a single utterance. Nobody is even assigned to the question.

The Layer That Is Still Missing

Step back and look at what has been built in the past eleven months:

  • Discovery: agents now browse, compare, and rank products on the consumer’s behalf.
  • Checkout: standardized across four competing protocols from OpenAI, Google, Microsoft, and Amazon.
  • Financing: now embedded in the agent itself, courtesy of the Synchrony plugin.
  • Loyalty: rewards and offers flowing through the same channel.
  • Returns: covered by protocol, and consumers treat return windows as their trust hedge.

Every step of the transaction is now agent-mediated. Zero steps of it are independently verified. The RTB House survey found that 42% of US millennials would let an agent buy within a $250 budget only if returns were guaranteed within seven days. Consumers have noticed the gap. They are pricing it manually, with the only tool they have: reversibility.

But reversibility is a post-hoc fix for a pre-purchase problem. It substitutes shipping hassle and reverse logistics for judgment. The cheaper mechanism, for everyone, is verification before the transaction: an independent read on whether the product’s reviews are authentic, whether its rating survives the removal of manipulated feedback, and whether its quality justifies the price before financing is attached to it.

That is the layer we build. GoBuy’s Smart Score rates products 0-100 based on review quality, not review quantity, filtering fake and incentivized reviews before they can inflate a recommendation. Our MCP server at gobuy.ai/api/mcp lets any AI agent, including ones considering a “six months, no interest” offer, check what it is about to finance before the financing API is ever called. The Chrome extension puts the same trust panel directly on Amazon pages, for the shoppers who still do the final click themselves.

The Question the Industry Skipped

The Synchrony-OpenAI announcement ends with the phrase “trusted, AI-powered commerce.” Trustworthy agentic commerce is achievable. But the word has to mean more than “secure payments and responsible lending.” A financed purchase of a manipulated product is a perfectly secure transaction that still defrauds the consumer. It just does so with interest attached.

The industry has now standardized everything about the transaction except the truth of the product. Until an independent trust layer sits between the recommendation and the checkout, every dollar of agent-mediated volume, consumer projections run to $15 trillion by 2028 on the B2B side alone, will flow through the least verified decision point in the history of retail.

Before your agent checks out, make it check first. GoBuy at gobuy.ai, and agent documentation at gobuy.ai/agent-docs.

Sources: Synchrony press release (PR Newswire), Harvard Business Review, “Algorithmic Shopping Is Here. Is Your Company Ready?”, Gartner predictions for 2026 and beyond, Synchrony Financial corporate history.