AI Hallucination and Oversight

Generative AI has the potential to improve efficiency, accelerate insight generation, and support decision-making in real-world evidence research, but also introduces risks requiring oversight and validation. Regulatory and health technology assessment agencies have already published position statements on generative AI use in real-world evidence. Hallucination, where a model generates coherent but inaccurate or unsupported content, is identified as a central concern for generative AI in real-world evidence contexts. The review aims to provide evidence to support future benchmarking standards, regulation, governance, and transparent reporting guidance for generative and agentic AI in real-world evidence generation.