Key findings
- Shopping starts with AI. Nearly 61 million consumers begin their research for online retail purchases with AI. Average AI-assisted spending rose 17% between the June and August surveys.
- Amazon closes what AI opens. Six in 10 respondents who used AI for shopping completed their latest AI-assisted purchase on Amazon. Price, shipping and payment convenience help explain destination choices.
- Research moves more readily than checkout. Fifty-nine percent of consumers who used AI for shopping or discovering products say they now research on AI what they previously searched for on Amazon. Yet only 27% say they buy from Amazon less often.
- An agent can inherit the shopper’s habits. With the item, price, shipping and returns held equal, 40% would direct an autonomous assistant to a large marketplace, the largest single group
AI’s shopping audience is growing even as its broader consumer audience holds steady. Between June and August, the share of consumers who had used generative AI for personal tasks in the preceding year changed little, from 57.6% to 58.3%.
Product discovery tells a different story. The share of consumers who used AI to discover products at least once in the preceding year rose from 22.9% to 25.8%, reaching 67.8 million consumers. The narrower group that treats AI as its main research starting point for retail purchases grew from 21.4% to 23.2%, reaching 60.9 million.
Trying AI once and making it the first stop are different behaviors. The latter puts an assistant in a position to help shape the shortlist before a shopper visits a store’s website.
Spending adds another dimension. Among AI shoppers, estimated average total spending on AI-assisted retail purchases over the preceding three months rose from $889 in June to $1,039 in August, a 17% increase.1 These are spending totals for the period, not average order values.
The findings suggest a deeper shopping role within a broadly stable AI audience. Companies can gain more influence over purchases by giving existing users stronger reasons to return for advice.
Familiar checkout advantages still close the sale
Among respondents reporting their most recent AI-assisted purchase, 59.7% completed it on Amazon. Big-box and department stores followed at 19.5%, while brands’ own websites or apps accounted for 6.3%.
Across purchase destinations, 27.5% named the best price or a discount as their most important reason for buying there. Free or fast shipping followed at 17.3%. Trust in the destination’s handling of payment details drew 9.1%, while 8.1% cited habit.
Just 2.5% named a direct link from the AI tool as the main reason. An assistant can help consumers choose a product without determining where they buy it. The merchant still has to make the transaction attractive and dependable.
AI assistance also spans categories with different research needs. Clothing, shoes and accessories accounted for 27.7% of the most recent purchases, followed by electronics and technology at 21.8%. Together, the two categories represented nearly half the reported purchases. Appliances contributed another 7%.
For retailers, the opportunity extends beyond appearing in an AI recommendation. A competitive offer and a familiar payment experience can convert research that happened on another platform. Amazon’s strong showing illustrates how an established checkout relationship can retain value even when the first shopping conversation happens elsewhere.
Displacement happens unevenly
Consumers themselves describe the split between research and purchase. Among those who used AI in the past year for product discovery or shopping and purchasing, 59.4% agreed that they now use AI to research products they used to search for on Amazon. Some 57.9% agreed that they usually buy on Amazon even when AI helps them find the product.
The destination comparison reinforces that pattern. Among consumers who now start their research with AI, 41.7% said Amazon was where they most often bought before AI was available. Across all respondents, 49.3% said Amazon is where they generally buy online today, and 59.7% of consumers whose recent online purchases involved AI completed the most recent one on Amazon. The difference between the first and third figures is 43% in relative terms, which underscores that AI is displacing other research tools without displacing the purchase destination.
These are self-reported comparisons, including a remembered baseline. They don’t establish that AI caused Amazon’s gains or measure changes in total sales.
There is nevertheless meaningful movement elsewhere. Some 44.7% agreed that using AI had led them to buy from online stores they’d never used before. Another statement, that AI had changed where they buy, drew 39.9% agreement. And 27.2% said they have bought from Amazon less often since they began using AI for shopping.
Discovery can open doors for unfamiliar merchants, while familiar destinations retain the stronger overall position. For Amazon, the results suggest greater competitive exposure in product search and advertising than in checkout. For other sellers, a recommendation creates an introduction they still must turn into a customer relationship.
Consumers give agents instructions and preferences
The agentic checkout exercise tests a possible next step. Consumers considered an identical item offered at the same price, with the same shipping and returns, across three types of sellers. They then indicated where they’d want an autonomous AI assistant to buy it.
A large online marketplace led at 40.2%. A department store or large retailer drew 21.8%, while a brand’s own website attracted 16.1%. Another 21.8% had no preference and would let the assistant choose.
The marketplace drew the largest share, but roughly 60% chose another destination or left the decision open. That creates room for competing services. It also shows that delegating a purchase doesn’t necessarily mean surrendering preferences about the seller.
In a related exercise on destination choice, consumers who preferred one option cited reliable returns, selection and habit despite the scenario’s equal terms. Those responses suggest that past experience shapes confidence beyond the stated offer.
Marketplace shoppers also identified reasons to buy directly from a brand or store. A lower price or exclusive discount appealed to 53.9%, free shipping as fast as the marketplace’s to 49.9% and equally easy returns to 35.2%. An AI assistant handling checkout appealed to 12.1%.
Only 11.3% said nothing would move them. The opening is real, but the strongest incentives remain commercial. Making checkout automatic is less persuasive than improving the deal.
Implications and opportunities
AI companies can build on the role consumers already value: helping them make a purchase decision. Clear comparisons and accurate prices can earn repeat use. For buying agents, preserving a shopper’s preferred store and asking permission before changing sellers may be as valuable as finding an alternative. This wave doesn’t measure how many purchases consumers complete inside AI platforms.
Brands and retailers can pursue discovery and conversion separately. Product information needs to be useful to an assistant, while the offer needs to justify choosing the store. Marketplace sellers should also keep their identities clear: 64% of all respondents say they always or sometimes check whether an Amazon item comes from Amazon itself or a third-party seller.
Payment providers can help retailers carry familiar conveniences into new shopping experiences. Saved wallet access is one example: 23.9% of marketplace shoppers said it could encourage a direct purchase. The opportunity is to make payment easy while preserving authorization and confidence in the seller. Retailers will still need to compete on price, delivery and service.
Conclusion
The next stage of AI shopping will test how much of the journey consumers want to delegate. They can hand an assistant the research and still retain a strong preference about who fills the order. Companies that connect useful advice with a compelling offer and dependable service will have the clearest opening to win the purchase.
Read More
PYMNTS Intelligence is the leading provider of information on the consumer trends driving innovation in consumer finance, digital payments and financial inclusion. To stay up to date, subscribe to our newsletters and read our in-depth reports.
Methodology
“From Prompt to Purchase: Who Wins the AI Shopper?” draws on the monthly Consumer AI Benchmark Report. This edition uses the August 2026 survey of 2,191 U.S. consumers, balanced to the U.S. census by age, gender, income and region and weighted to represent 262.3 million consumers. Discovery trends compare the June, July and August waves. August wave was fielded from August 10–13, 2026.
Pre-AI comparisons rely on recalled behavior. Spending averages use midpoint estimates from response bands and cover the preceding three months. Percent changes are relative differences, not percentage-point changes.
1. June and July surveys asked about all retail purchases, while the August survey asked only about online purchases.↩
