Featured in Business InsiderDakota Shane Nunley talks about a new study he and his team created, and why AI isn’t ready to do your holiday shopping yet.Read the story ↗
Dakota/Shane
Research

Original research.

Studies from the research program I built at Product.ai (formerly Demand.io). They show AI shopping assistants getting product facts wrong, shoppers double-checking what AI tells them, and retailers pulling back on promo codes. Every study publishes its method alongside its findings.

86% of questions produced a factual conflict across ChatGPT, Claude, Gemini, and Perplexity
· Product.ai · Sean Fisher and Dakota Shane Nunley

We asked four AI engines the same 220 shopping questions

When an engine got a price wrong, it missed by a median of $300.

Method220 shopping questions across nine product categories, asked of each engine’s free and paid tiers, five runs each: 8,794 answers.

Covered by
Read the study ↗
43% of U.S. online shoppers used an AI assistant to research a purchase in the past 90 days
· Product.ai · Product.ai Research

The 2026 Trust in AI Commerce Report

86% of them checked the AI’s pick against another source before buying.

MethodSurvey of 1,463 U.S. online shoppers.

Covered by
Read the study ↗
45% drop in monthly promo codes from the October 2023 peak, across 80 national retailers
· SimplyCodes, Product.ai’s coupon platform · Sean Fisher

We tracked 13,549 promo codes

Method13,549 promo codes from 80 retailers, September 2023 through January 2026.

Covered by
Read the study ↗

How I design a study so it gets cited: What Is a Canonical Stat? →