What happened
- Product.ai published a new edition of its AI shopping trust survey on October 14. It questioned 1,493 people in the United States between August 27 and September 1.
- Among the 550 people who had used an assistant to research a purchase, 89.9% said they checked its recommendation against another source before buying. The June 2026 edition recorded 85.7%.
- 21.6% of that group said they had let an assistant complete a purchase. The report also asks what spending authority participants would accept.
Why it matters
- For brands and retailers, appearing in an assistant’s answer is not the same as making a sale. A user may check price, product features and reputation elsewhere before deciding.
- The report comes from a company that sells a commerce verification layer. Its survey does not represent Chilean shoppers and relies on self-reported answers; it is a signal for testing local behavior, not a national statistic.
- Teams that manage product catalogs can reduce friction by keeping information consistent and verifiable. A bad recommendation can send a shopper to a different purchase: the study says six in ten respondents who caught errors changed their plans.
The number
89.9%of surveyed AI-assisted shopping researchers said they verified a recommendation against another source.https://www.prnewswire.com/news-releases/productai-survey-finds-four-in-five-ai-shoppers-havent-let-an-ai-assistant-buy-for-them-nine-in-10-check-its-advice-before-purchase-302907868.html
Context
The survey combines 1,493 shoppers, 550 recent assistant users and different subgroups depending on the question. Product.ai publishes a codebook and methodology, but fieldwork was conducted in the United States. “Nine in ten” describes respondents who used AI to research a purchase, not all shoppers or all countries.
What’s next
- The report anticipates more assistant use during the U.S. holiday shopping season; it does not confirm the same trend in Chile.
- No date was announced for the next periodic survey.
Bottom line
An automated recommendation still needs a second opinion. For commerce, that second opinion is often a page the shopper already knows.
Sources
- Product.ai’s AI commerce trust survey, distributed by PR Newswire.
- GoDaddy, Google, Cloudflare and Infoblox’s agent transaction architecture.
Written by Mamífero. Edited by Rodrigo Cornejo. This draft was assisted by AI and requires human editing before publication. See how we select and verify these notes.
