Competitive research · Q2 2026
Retail AI Competitive Benchmark
How Amazon, Walmart, and Kroger's AI shopping assistants compare on the same real shopping tasks
- AI shopping assistants, same prompts
- 3
- points between Amazon (11.80) and Walmart (11.35)
- 0.45
- Kroger's acceptance score, the lowest in the study
- 5/15
- Walmart's credibility score, behind both rivals
- 9/15
The opportunity: help Kroger differentiate its AI shopping assistant against the two competitors customers already compare it to.
Amazon, Walmart, and Kroger each ship an AI assistant for shopping. To compare them fairly, the study uses a heuristic: a weighted scorecard that turns "which one feels better?" into specific, repeatable criteria. The three assistants were then scored on the same real tasks, from planning dinners to checking food safety to revising a list on a budget. No assistant won everywhere, and the gaps showed where Kroger was leaving value on the table.
My role: research lead and sponsor
I sponsored and led this study from start to finish, and co-authored it with a research partner:
- Commissioned it and got it funded, making the case for a head-to-head look at how Amazon, Walmart, and Kroger's AI assistants perform on the same tasks
- Led the direction of the study, from the tasks the assistants were given to what the findings needed to answer
- Informed the heuristic decisions, shaping the six dimensions and the one-to-five scoring that turn "which one feels better?" into repeatable criteria
- Used the results as a reference marker for other studies. Where our other research pointed in a similar direction, the findings were not far apart, which gave both bodies of work confidence in where to go next
The results below are what that heuristic surfaced.
The study
- 3 assistants: Amazon Alexa, Walmart Sparky, and the Kroger Shopping Assistant
- 6 dimensions of the experience, each weighted by how much it matters
- 3 prompts per dimension, identical across assistants
- Every rating is 1 to 5, built from five one-point criteria so scores trace back to specific behaviors
