Researchers find agent pricing follows whatever prices you show it

Researchers studied shopping agents built on large language models, the AI systems that read and write text. Agents that could see competitor prices held prices above the market floor. Agents that could not see competitor prices pushed prices down to that floor. The deciding factor was the information a designer chose to supply, not the AI model itself. The paper treats that information-supply decision as a governance choice, not a setup detail. An operator who shows an agent live market prices is choosing a pricing regime. For anyone designing a purchasing or pricing product, the information you give an agent at the start shapes the market outcome. Regulators may care about that outcome. For design leadership, the finding makes information-scoping a policy decision. Handing a task to an agent is not the end of the human role. The designer who assembles the agent's starting information feed makes a choice that carries through each action the agent takes. Governance starts at the information-supply stage, before the agent runs.