What Is Your AI Agent Buying? Evaluation, Implications and Emerging Questions for Agentic E-Commerce
Amine Allouah, Omar Besbes, Josué D Figueroa, Yash Kanoria, Akshit Kumar
2025-08-06
Summary
This paper talks about ACES, a testing environment that studies how AI shopping agents behave when they shop in a fake online marketplace, showing how these agents react to things like product placement, price, reviews, and advertisements.
What's the problem?
The problem is that as AI agents start shopping for people, we don’t really understand how their decisions are influenced by different factors in online stores, and this raises questions about fairness, competition, and how sellers should design their strategies.
What's the solution?
ACES solves this by creating a controlled, fake marketplace where AI agents can shop, and researchers can change things like where products appear, their ratings, prices, and ads, to see how these changes affect what the AI agents buy and why.
Why it matters?
This matters because understanding AI shopping behaviors helps businesses and platforms design better online stores and fairer markets, and it raises important questions about how AI might shape the future of e-commerce.
Abstract
ACES, a sandbox environment, studies AI agents' shopping behavior in a mock marketplace, revealing position effects, sensitivity to sponsored tags, endorsements, prices, ratings, and reviews, and highlighting implications for seller strategies and platform design.