A Survey on LLM-based Multi-Agent AI Hospital

AI hospitals use agent-driven multi-agent systems based on large language models (LLMs) to automate and optimize medical workflows, enabling intelligent agents to understand, reason, and assist in complex medical tasks. Although AI-driven healthcare applications are developing rapidly, research remains fragmented across various scenarios. This survey carefully analyzes 72 studies on LLM-based medical agents published between 2023 and 2025, provides a comprehensive review of AI hospitals, and systematically introduces a structured taxonomy to categorize its core components and applications. Additionally, we explored the key challenges associated with the core components of AI hospitals, including agent roles, interaction patterns, reasoning mechanisms, memory management, and tool integration. Finally, we explore how to further develop the AI hospital into a more meaningful research and practice platform that supports medical simulation, complex problem-solving, evaluation, and synthetic data generation. This, in turn, will accelerate progress in clinical reasoning, decision support, and AI-driven medical innovation, highlighting the crucial role of AI hospitals as the foundational framework for AI-powered healthcare ecosystems. By providing a structured perspective, this survey bridges AI and healthcare research, providing a roadmap for strengthening interdisciplinary collaboration and the practical applicability of AI hospitals.