I use a large language model to generate synthetic survey responses to different bank run scenarios. By assigning the model with specific demographic characteristics, I simulate diverse populations of agents, and assess their responses to varied conditions and communication-based interventions related to a bank run. The simulated responses, reflecting trends in bank deposit withdrawals across demographic categories, align with existing empirical studies. This novel application offers a cost-effective method to analyze hypothetical scenarios using simulated data, and highlights the potential of language models in providing insights into human behavior in economic context.