What changed when the agent acted?
Compare the brands that led in recommendation mode with the brands that led when the agent chose a product or tried to add it to a cart.
KNWN Research · October 2026
Do AI shoppers buy the brands they recommend?
We gave an AI shopping agent the same task in different modes: recommend the options, choose a product, or add it to a cart. The brand it recommended wasn’t always the brand it chose.
- 400
- shopping runs
- 20
- shopper scenarios
- 5
- product categories
AI Shopper Benchmark
By Emily Richardson & Khazar Ayaz
Inside the report
From a recommendation
to a decision.
A closer look at how one AI shopping agent navigated real retail sites, and why visibility alone didn’t always lead to a choice.
Where did the journey stop?
Trace the gaps between finding a product, confirming it was buyable, comparing the options, and reaching a cart.
What should brands investigate?
Understand what these observations suggest for product pages and shopping flows, alongside the limits of the study.
A benchmark, with boundaries.
The study ran on September 13–14, 2026, across footwear, consumer electronics, home and kitchen, personal care, and office, fitness and travel. These were repeated runs of one AI agent, not a sample of human shoppers. No purchases were completed. The results are not conversion rates or market shares.
Download the full methodology