Magentic Marketplace: An Open-Source Environment for Studying Agentic Markets
Gagan Bansal, Wenyue Hua, Zezhou Huang, Adam Fourney, Amanda Swearngin, Will Epperson, Tyler Payne, Jake M. Hofman, Brendan Lucier, Chinmay Singh, Markus Mobius, Akshay Nambi, Archana Yadav, Kevin Gao, David M. Rothschild, Aleksandrs Slivkins, Daniel G. Goldstein, Hussein Mozannar, Nicole Immorlica, Maya Murad, Matthew Vogel, Subbarao Kambhampati
2025-10-31
Summary
This paper investigates how well AI agents, specifically large language models, perform when acting as both consumers and businesses in a simulated marketplace, and what issues arise as these agents become more common in online economic activity.
What's the problem?
Currently, most research on AI agents in economic situations focuses on very simple scenarios, like one-on-one negotiations. Real-world markets are much more complex, involving many different agents, a wide variety of tasks, and ongoing interactions. We don't really understand how these AI agents will behave when placed in a realistic, dynamic market environment, and whether they'll be fair to users or lead to efficient outcomes.
What's the solution?
The researchers created a simulated marketplace called Magnetic-Marketplace where AI agents can act as shoppers (Assistants) and sellers (Services). This allows them to safely study how the agents interact, how much value they create, if they show biases in their behavior, and if they can be easily tricked. They then ran experiments with current AI models to see how they performed in this environment, focusing on things like how well they find good deals and how quickly they respond.
Why it matters?
This research is important because as AI agents become more involved in our everyday purchases and economic decisions, we need to understand their potential downsides. The study found that while AI agents *can* perform well under ideal conditions, their performance drops significantly as the market gets larger, and they tend to favor the first offer they see, giving an unfair advantage to faster responders over those offering better quality. This helps us design better, fairer, and more efficient AI-powered marketplaces in the future.
Abstract
As LLM agents advance, they are increasingly mediating economic decisions, ranging from product discovery to transactions, on behalf of users. Such applications promise benefits but also raise many questions about agent accountability and value for users. Addressing these questions requires understanding how agents behave in realistic market conditions. However, previous research has largely evaluated agents in constrained settings, such as single-task marketplaces (e.g., negotiation) or structured two-agent interactions. Real-world markets are fundamentally different: they require agents to handle diverse economic activities and coordinate within large, dynamic ecosystems where multiple agents with opaque behaviors may engage in open-ended dialogues. To bridge this gap, we investigate two-sided agentic marketplaces where Assistant agents represent consumers and Service agents represent competing businesses. To study these interactions safely, we develop Magentic-Marketplace-- a simulated environment where Assistants and Services can operate. This environment enables us to study key market dynamics: the utility agents achieve, behavioral biases, vulnerability to manipulation, and how search mechanisms shape market outcomes. Our experiments show that frontier models can approach optimal welfare-- but only under ideal search conditions. Performance degrades sharply with scale, and all models exhibit severe first-proposal bias, creating 10-30x advantages for response speed over quality. These findings reveal how behaviors emerge across market conditions, informing the design of fair and efficient agentic marketplaces.