Trust in AI Commerce Report (Product.ai)
Original study- Publisher
- Product.ai
- Evidence:
- Self-reported behaviour
- Vendor interest:
- High
- Sample:
- Survey of 1,493 U.S. adults who shop for products, online or in stores, from a general-population panel; unweighted with quotas on age, gender, and region. Trend comparisons use 1,247 online shoppers; AI-user findings mostly use 550 respondents who used AI for product research in the past 90 days.
- Methodology
- Disclosed
- Geography
- US
- Data period
- 27.08.2026 – 01.09.2026
- Published
- 14 Oct 2026
Summary:
This report from Product.ai, a company that builds verification products for commerce, tracks how U.S. shoppers use, trust and check AI product recommendations. It is the second edition of a recurring survey (Edition 2, published October 14, 2026) and repeats 12 questions from the June 2026 edition word for word to measure what changed.
The survey covered 1,493 U.S. adults who shop online or in stores, on a general-population panel, from August 27 to September 1, 2026. Quotas balanced age, gender and region; results are unweighted. Comparisons with the first edition use the 1,247 online shoppers, the same group the first edition sampled. Most AI findings come from the 550 respondents who used an AI assistant to research a purchase in the past 90 days.
Key findings:
• 41.5% of online shoppers used AI to research a purchase in the past 90 days, against 42.6% in June. The report treats adoption as unchanged.
• 89.9% of AI users verified an AI recommendation before buying, up from 85.7%. It is the only repeated measure that rose.
• 89.9% of shoppers who verify have caught AI wrong at least once, and 60.3% of those changed course the last time: they skipped the purchase, switched products or stopped using AI for that category.
• 36.7% of verifiers check online customer reviews first; 21.9% go to a search engine.
• 21.6% of AI users say they let an assistant complete a purchase start to finish. Completion is 28.8% among those who always verify and 8.8% among those who never do.
• Gemini (63.5%) and ChatGPT (60.0%) lead research use, and 64.0% of AI users rely on two or more assistants.
• 37.5% of AI users would let an agent spend nothing without asking, and 43.1% would not give it a payment method.
• AI ranks last of seven research sources on trust, at 4.12 out of 7, the same as in June. Friends and family lead at 5.34.
• A hidden sponsorship costs more trust than a labeled one: 50.4% vs 39.8% would trust the assistant less.
• 75.1% knew or suspected AI can repeat brand marketing as fact, yet 48.7% agree AI shopping answers are unbiased.
• Only 14.0% of shoppers say “verified” means an independent check; 14.9% don’t know what it means.
Margin notes:
This is a vendor report: Product.ai sells verification products, and the findings on checking AI answers and on the word “verified” sit close to what it sells. All behavior is self-reported, so verification (a socially approved answer) may be overstated, and “completed a purchase” depends on how each respondent read the question. The sample is U.S.-only and unweighted, and two editions make a comparison, not a trend. Subgroups by age, region, baby-product buyers and the 119 shoppers who completed an AI purchase are small, and the report marks them as directional. The assistant trust ranking comes from the survey platform’s standard report and can’t be recomputed from the public data; the report itself notes it tracks familiarity as much as trust.