Most research on agentic commerce looks at the buyer's side: how well an AI agent can browse, compare, and purchase. A new paper from arXiv argues that the merchant's side is just as important, yet largely unmeasured. The authors introduce a readiness benchmark that evaluates whether a website actually lets an agent in, publishes clear terms, and supports the steps needed to complete a transaction.
The benchmark is structured around three stages: getting in, deciding, and proving. Getting in covers access and technical compatibility; deciding involves whether the site presents enough information for an agent to make choices; proving concerns verification and confirmation of the purchase. By scoring merchants on these dimensions, the benchmark aims to identify where websites fall short for autonomous shoppers.
The paper does not claim to solve these problems, but it provides a framework for measuring them. It complements existing agent-side benchmarks rather than replacing them. The authors suggest that both sides need to improve before agentic commerce becomes practical at scale.