Generative artificial intelligence (AI) is doing something that years of retail competition struggled to accomplish. It is teaching consumers to start shopping somewhere other than Amazon.
The catch? AI just isn’t teaching them to stop buying there. And that disconnect may be one of the most consequential early signals in agentic commerce.
New PYMNTS Intelligence data from the report, “Will the 2026 Shopping Season Go Agentic?” shows Amazon accounts for 29% of consumers’ online retail research starts but 49% of their main purchase completions. Among consumers’ most recent AI-assisted purchases, Amazon’s share rises to 59%. The measures come from separate distributions rather than a tracked funnel, but the direction is notable and shows how AI is disrupting discovery much faster than it is disrupting transaction capture.
For retailers, AI platforms and payments companies racing to own agentic commerce, that’s the more interesting battle. The competitive advantage may not belong to whoever builds the smartest shopping agent. It may belong to whoever already owns everything the agent needs once the shopper says yes.
AI Is Unbundling Search From Checkout
The emerging agentic shopping divide is not simply between human commerce and autonomous commerce. It is between AI controlling discovery and established commerce infrastructure controlling execution.
AI shopping has already crossed into the mainstream. Roughly 132 million U.S. adults have bought something with AI assistance, as shared by PYMNTS Intelligence, while 58.4 million now begin online retail research using an AI tool. Yet only 62.9 million would let an agent both shop and buy for them, and just 36.8 million would allow the AI application to hold their card.
AI is becoming a powerful new purchase research layer. But research isn’t checkout.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
That distinction explains a particularly counterintuitive finding: 57.5 million AI-assisted shoppers say they now research through AI things they previously searched for on Amazon, yet 54.5 million still buy on Amazon even when an AI tool found the product.
Read the report: Will the 2026 Shopping Season Go Agentic?
AI can search the internet, compare products and compress dozens of tabs into one answer. But once a shopper wants to transact, Amazon can offer something a recommendation engine cannot instantly manufacture: stored payment credentials, known identity, familiar delivery expectations, loyalty benefits and an established returns process.
The PYMNTS Intelligence report calls this a “ladder” of delegation rather than a switch. That distinction matters because willingness to let an agent both shop and buy declined from 32% in September 2025 to 24% in August 2026. Consumers aren’t necessarily rejecting agents. They appear to be distinguishing between tasks according to their perceived risk.
The winning agentic checkout, then, could look less like handing an AI unrestricted purchasing power and more like issuing it a tightly scoped corporate card: spend up to a defined amount, at approved merchants, under specified conditions, and come back for authorization when the rules are exceeded.
Eighty-seven percent of merchants surveyed by PYMNTS Intelligence for the “Will the 2026 Shopping Season Go Agentic?” see agentic commerce as an opportunity, 90% have begun product and catalog work, and businesses report spending an average 2% of annual revenue preparing for it. Yet only 23% say they are very or fully prepared to integrate with AI platforms or agent networks.
That turns the central problem of agentic commerce from “Can AI buy something?” into “Who defines, authenticates and enforces the AI’s authority to buy it?”
At PYMNTS Intelligence, we work with businesses to uncover insights that fuel intelligent, data-driven discussions on changing customer expectations, a more connected economy and the strategic shifts necessary to achieve outcomes. With rigorous research methodologies and unwavering commitment to objective quality, we offer trusted data to grow your business. As our partner, you’ll have access to our diverse team of PhDs, researchers, data analysts, number crunchers, subject matter veterans and editorial experts.
Add as Preferred Source
