In ChatGPT, 84% of Buying Decisions Came From Product Cards, ReFiBuy Study Finds
New behavioral study shows why getting recommended in answer engines is only the start: brands and retailers need to
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
New behavioral study shows why getting recommended in answer engines is only the start: brands and retailers need to compete for visibility and position in the product and offer cards where buying decisions take shape
RALEIGH, NC / ACCESS Newswire / October 6, 2026 / Product cards dominated buying decisions in a new behavioral study of AI shopping commissioned by ReFiBuy and conducted by Clickstream Solutions. In ChatGPT, 84% of final product choices came from product cards, the visual listings that bring product images, prices and key details into the AI conversation. Rather than relying on surveys or traffic data alone, the study observed what shoppers compared, rejected and chose during shopping tasks, with spoken commentary revealing the reasoning behind those decisions.
The study, In AI Shopping, the Product Card Is Your Storefront, recorded 40 U.S. participants completing 224 shopping tasks across six product categories in ChatGPT and Google AI Mode. The recordings captured how shoppers compared products and chose between offers, including product cards they considered but never clicked.
Key Findings
- 84% of final product choices in ChatGPT came from a product card. If a product misses the product card set, it misses the comparison where most choices were made.
- 75% of shopping tasks started with product-card engagement. Shoppers went to the product cards first, making the image, title, price and other displayed product data an early part of the buying decision.
- 43% of choices went to the first product card, more than any other position. Position pays, making the top spot worth treating as a priority.
- 76% of offer choices went to the first offer card. Position pays again at the offer level, where price, availability and delivery shape the next step.
“The Agentic Commerce user behavior study proved our thesis that, like the Amazon Buy Box, product cards on ChatGPT are critical to winning the consumer’s transaction,” said Scot Wingo, co-founder and CEO of ReFiBuy. “84% of final product choices in ChatGPT came from a product card. Being mapped correctly to a product card and appearing as high as possible in the offer card list are absolutely critical to optimizing for ChatGPT.”
“When an AI assistant showed two or more product cards, shoppers chose the first one 43.4% of the time,” said Eric Van Buskirk, founder of Clickstream Solutions. “Chance would put that at about 29%. Nearly all of the advantage goes to the top slot. The second card was chosen less often than chance would predict.”
What Brands and Retailers Can Do Now
For brands investing in answer engine optimization (AEO) and generative engine optimization (GEO), visibility is the start of the job. The work continues in the product and offer cards where products compete for the buying decision. ReFiBuy calls that work Agentic Commerce Optimization (ACO).
ReFiBuy turned the findings into eight actionable recommendations for brands and retailers, from getting products into the product cards and treating the first position as a priority to keeping prices and variants consistent across product cards, offer cards and product pages.
The full report also includes recorded session captures, what shoppers relied on when they made a choice, how they reacted to sponsored placements, and the price mismatches they caught on product cards while deciding.
The full study and all eight recommendations are at refibuy.ai/study.
About the Research
ReFiBuy commissioned the study. Clickstream Solutions designed it, coded the sessions and ran the analysis. Forty U.S. participants, recruited through Prolific and all existing ChatGPT users, completed 224 shopping tasks in six categories on ChatGPT and Google AI Mode in remote, unmoderated desktop sessions.
The headline finding draws on 104 ChatGPT tasks in watches, consumer electronics and bags, where participants were not told to choose from a product card. Position was observed, not randomized, so the findings show where choices concentrated rather than proving that position caused them. The study measured choices, not purchases. Full methods, task instructions and limitations are in the report.
About ReFiBuy
ReFiBuy coined Agentic Commerce Optimization (ACO) in 2025 and defined the operating model leading brands and retailers use to make catalogs perform in AI Shopping. Its Commerce Intelligence Engine is the first agentic-native, closed-loop platform built to put ACO into practice.
The platform evaluates, enriches, and distributes product data, then monitors how products are interpreted and chosen. Live agent behavior and shopper context feed back into the catalog, making it smarter as shopping changes across answer engines and retailer-hosted AI agents.
Founded by ecommerce veterans from ChannelAdvisor (now Rithum), Walmart, and MikMak, ReFiBuy is trusted by leading brands and retailers across categories. Learn more at refibuy.ai.
About Clickstream Solutions
Clickstream Solutions runs studies on how people use Google and AI search. Its private studies help clients discover how AI platforms shape awareness of their products compared with direct competitors. It also partners on published research, pairing its findings with clients’ industry expertise. Learn more at clickstream.cc.
Media Contact
Brian Chapman
VP of Growth
ReFiBuy
brian@refibuy.ai
SOURCE: ReFiBuy
View the original press release on ACCESS Newswire
Media gallery
