AI-generated expectations and market stability

Generative Artificial Intelligence (AI) is rapidly emerging as a new technology to form economic expectations, raising new concerns about market dynamics and stability. I develop a general-equilibrium experimental asset-pricing framework that enables systematic evaluation of the stabilizing implications of AI-generated expectations in the presence of shocks unrelated to fundamentals. Applying the framework to a set of AI models, I find that AI expectations induce price dynamics that are markedly more stable than human expectations in equivalent environments, yet systematically deviate from Rational Expectations. This stability proves robust to temporary price shocks in most settings, but I document cases in which it breaks down, depending on the combination of the AI model and its use. I introduce a simple indicator that summarizes the destabilizing tendency after shocks of each model– use combination. The framework can be applied to new AI models, alternative shocks, and endogenous adoption of AI forecasting technologies.