AI-powered shopping is moving faster than most merchants can keep up. ChatGPT, Perplexity, and Google's Gemini are all competing to help consumers buy products directly through conversational interfaces, but the infrastructure to actually make it work remains fragmented. AI platforms want unique integrations, protocols all look different, and merchants must continue to think about which standards to adopt while their core business gains traction in other areas.
PayPal positions itself as the infrastructure layer that connects everything.
At Money20/20, I spoke with Michelle Gill, executive vice president and general manager of PayPal's small business and financial services group, about the company's transition to agent commerce. The pitch: Merchants are drowning in new protocols like MCP servers, API variations, and different LLM integrations, and don't have time to decide which horses to bet on. Meanwhile, PayPal predicts that more than 40% of Americans will start discovering products with AI surfaces this holiday season. It's a discovery, not a purchase, but the behavioral change is already happening.
“New protocols are coming out every day,” Gill told me. “Sellers are trying to figure out how much time they should be spending with agents, but they're already reaching their limit because they don't have time to do anything.”
PayPal's solution: Becomes an abstraction layer between merchants and all AI shopping surfaces. One integration that propagates to ChatGPT, Gemini, Perplexity, and whatever comes next. Orders are routed back through existing infrastructure, rather than requiring separate coordination systems for each platform.
OpenAI’s recent announcement that PayPal’s wallet will be integrated into ChatGPT’s shopping experience shows the strategy in action. PayPal collaborated with OpenAI to create an open wallet protocol. They position themselves not as a destination to compete with these platforms, but as a neutral infrastructure that makes them all work.
What makes this interesting is the SMB angle. Gill's argument is that without a centralized infrastructure, large retailers will dominate early AI shopping integrations, leaving smaller merchants behind. The divide that already exists in digital advertising (those who spend the most on ads wins) could become even worse with AI commerce. But the economics change as discoverability moves from search rankings to conversational AI. Small sellers don't have to spend more money to gain recognition on AdWords than their competitors. AI agents display products based on suitability, not ad budget.
“If discoverability is not based on beating AdWords, then that gap is actually potentially neutralized,” Gill says.
Whether it works or not depends on how the AI platform builds its recommendation logic. Nothing will change if you continue to pay. If it's truly based on matching consumer needs and product suitability, then betting on PayPal's infrastructure makes more sense. Businesses don't need to own the shopping experience. You need to enable it everywhere it occurs.
The technical challenge is order management. Things like verifying inventory exists, processing transactions, calculating taxes, and shipping between surfaces where PayPal doesn't control the UI. They are currently building out their seller onboarding tools, starting with a demo for early partners. Gill sees this as a strategic priority at PayPal, alongside its core products. The company has cross-functional teams that reuse existing architecture for new use cases. The company is also working with Nvidia to reduce computing requirements, as general-purpose models consume expensive processing power for relatively simple shopping tasks.
Where PayPal believes it has an advantage is consumer trust. Gill points out similarities to the days of the original eBay. PayPal became the abstraction layer because consumers didn't want to give out their credit card details to random sellers in other states. Consumers may not want their full payment credentials stored in an AI agent, so PayPal is once again that layer.
According to Gill, the biggest risk in deployment is security. If trust breaks down early, everything comes to a standstill. The short-term reality is more mundane. Merchants need conversational search interfaces that match the consumer expectations set by major retailers. PayPal wants to provide that functionality without requiring engineering resources that many of these merchants don't have.
PayPal stock rose 9% following the agency's announcement. Whether that happens depends on merchant adoption, which in turn depends on whether consumer behavior is changing faster than ever before. However, 400 million consumers and 20 million merchants have already joined the network, with an additional 2 billion joining the network through international wallet integration. That distribution exists. The question is whether being an infrastructure rather than a platform owner is as important as PayPal thinks it is.
