TL;DR
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
PYMNTS Intelligence reports that Amazon accounted for 59% of consumers’ most recent AI-assisted purchases, while 29% of online retail research starts began on Amazon. The figures come from separate distributions, not a tracked shopping funnel, but point to a gap between AI-led product discovery and where consumers complete purchases.
Amazon accounted for 59% of consumers’ most recent AI-assisted purchases, according to new PYMNTS Intelligence data, even as shoppers increasingly use artificial intelligence tools to research products. The figures suggest that AI is changing where people look for products faster than it is changing where they buy them, though the report cautions that its research-start and purchase-completion figures are separate distributions, not a tracked customer funnel.
The report, “Will the 2026 Shopping Season Go Agentic?,” puts Amazon at 29% of consumers’ online retail research starts and 49% of their main purchase completions. For consumers’ most recent purchases made with AI assistance, Amazon’s share was 59%. These measures describe different groups and stages; they should not be read as showing that individual shoppers moved from an AI tool to Amazon in a measured sequence.
Other figures in the report illustrate the reach of AI-assisted shopping alongside limits on delegation. PYMNTS Intelligence estimates that about 132 million U.S. adults have bought something with AI assistance, while 58.4 million now start online retail research using an AI tool. A smaller 62.9 million would allow an agent to both shop and buy for them, and 36.8 million would let an AI application hold their card.
The data also capture overlap between AI research and Amazon purchases. The report says 57.5 million AI-assisted shoppers now use AI to research things they previously searched for on Amazon, while 54.5 million still buy on Amazon even when an AI tool found the product. PYMNTS presents these as evidence that discovery and transaction are not necessarily shifting to the same service.
AI Discovery Meets Amazon Checkout
The reported split matters to retailers, payment providers and AI companies competing to shape online shopping. An AI assistant may help a customer identify a product or compare options without becoming the place where the customer pays. In that case, the service that wins attention during research may not capture the transaction or the customer relationship that comes with it.
PYMNTS points to Amazon’s established shopping infrastructure as one explanation for its purchase share: stored payment credentials, familiar identity and delivery expectations, loyalty benefits and returns. Those are factors identified in the report, not proof of why each shopper chose Amazon. They do, however, help explain why an AI recommendation alone may not remove the friction or perceived risk of buying through a new service.
The broader contest is over who can set and enforce an agent’s authority to spend. The report describes a “ladder” of delegation: consumers may accept help with research but hesitate to let software complete purchases or hold payment credentials. That points toward shopping tools that operate within clear spending limits and approval rules, rather than assuming users will hand over unrestricted buying power.
Products Worth Considering
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Dell XPS 13 9350 Copilot+ Laptop, 13.4-inch 120Hz Display, Intel Core Ultra 7 256V, 16GB RAM, 512GB SSD, Ubuntu Linux 24.04 LTS
Dell 24 Plus Monitor S2425HSM – 23.8 FHD (1920×1080) 144Hz 1ms, 2 x 3W Speakers, HDMI, Height/Tilt/Pivot/Swivel Adjustability, AMD FreeSync
As an affiliate, we earn on qualifying purchases.
A Gradual Shift in Shopping Tasks
The report describes agentic commerce as a developing form of shopping in which AI tools can research, recommend and potentially transact for consumers. Its figures suggest that these functions are not advancing at the same pace. AI-assisted research has reached a sizable group, but willingness to authorize an agent to both shop and buy remains more limited.
In PYMNTS Intelligence’s figures, the share of consumers willing to let an agent both shop and buy fell from 32% in September 2025 to 24% in August 2026. The report interprets this as a sign that consumers may distinguish among tasks according to their perceived risk. It does not establish why the measure declined, or whether the change reflects lasting attitudes, conditions in the survey, or other factors.
Businesses are preparing for the possibility that AI agents will become a more common route to customers. PYMNTS says 87% of surveyed merchants see agentic commerce as an opportunity, 90% have started product and catalog work, and companies report spending an average of 2% of annual revenue on preparation. Yet only 23% say they are very or fully prepared to integrate with AI platforms or agent networks, according to the report.
“The report describes consumer willingness to delegate shopping as a “ladder” rather than a switch.”
— PYMNTS Intelligence report, “Will the 2026 Shopping Season Go Agentic?”
Limits of the Purchase Data
The reported 29% share of research starts, 49% share of main purchase completions and 59% share of recent AI-assisted purchases are not a single tracked funnel. The source does not provide a common sample path showing how many shoppers who began research with AI later completed a purchase on Amazon. The figures therefore indicate a difference across reported measures, not a direct conversion rate or proof that AI caused a particular retailer choice.
The source material does not specify the survey field dates, sample sizes, question wording or margin of error for each measure. It also does not break down the 59% figure by product category, AI platform or transaction value. The reported merchant readiness and spending figures describe survey responses and should not be treated as a complete measure of the retail sector.
It remains unclear whether shoppers will become more comfortable authorizing AI agents to pay, or whether they will continue using AI mainly for discovery and then return to established retailer checkouts. The report identifies consumer control and merchant integration as concerns but does not establish which platforms or payment systems will resolve them.
Merchant Readiness and Agent Controls
The report points to catalog and product-data preparation as one area merchants are already addressing, while the relatively low share reporting strong integration readiness signals that work remains. For shoppers, the next practical test will be whether AI services can connect recommendations to reliable checkout without asking users to surrender more payment authority than they are comfortable granting.
Further reporting would need comparable, dated measures that follow shoppers from research through purchase to show whether AI-led discovery changes where transactions are completed. The supplied report does not announce a specific next release or implementation deadline. For now, its findings suggest that retailers and technology providers are competing not only to influence product choice, but also to earn authorization at the point of payment.
Key Questions
What does the 59% figure measure?
PYMNTS Intelligence says Amazon accounted for 59% of consumers’ most recent AI-assisted purchases. The source does not provide a category breakdown or establish that the figure measures all online purchases.
Does the report show that AI shoppers were sent to Amazon?
No. The report says the research-start and purchase-completion figures come from separate distributions, not a tracked funnel. It does not show the path of each shopper from discovery to checkout.
How many people use AI for online shopping research?
PYMNTS Intelligence estimates that 58.4 million U.S. adults now begin online retail research using an AI tool. The report also estimates that about 132 million U.S. adults have bought something with AI assistance.
Are consumers ready to let AI make purchases for them?
The reported measures suggest differing levels of comfort. The source says 62.9 million consumers would let an agent both shop and buy, while 36.8 million would allow an AI application to hold their card. It also reports that willingness to delegate both shopping and buying declined from 32% in September 2025 to 24% in August 2026.
Source: pymnts.com
