Current Applications, Research Gaps, and Future Perspectives in Design of Resilient Patient-Facing Generative AI Applications: A Rapid Review

Generative AI (GenAI) technologies, such as ChatGPT are increasingly being used by patients for healthcare applications, raising alarms about their reliability, safety, and appropriateness for patient care use cases. This rapid review synthesizes current research on the evaluation of patient-facing generative AI-based solutions in healthcare settings. We hope to summarize the scientific literature on emerging use cases, highlighting risks, mitigation strategies, and research gaps for future evaluation. This study aims to evaluate studies of generative AI applications with direct patient involvement. We conducted a rapid review using adapted PRISMA 2020 guidelines, searching in PubMed for studies published between January 2023 and January 2025. Articles were screened for inclusion and exclusion criteria, focusing on the evaluation of patient-facing applications. Our review resulted in 25 full-text studies. Potential applications include patient question answering, patient education material development, patient-friendly clinical report generation, patient self-management, mental health chatbots, and patient-reported outcome data collection. While early studies demonstrate promising outcomes such as increasing readability and patient comprehension, significant concerns persist around misinformation, hallucinations, and privacy concerns. Mitigation strategies include tailoring patient education on prompt design, retrieval-augmented generation, and transparency mechanisms. Key open questions remain about preserving privacy while incorporating patient preferences, building trust, and ensuring equitable access, especially for lower literacy populations. Very few studies focus on the evaluation of patient perspectives on generative AI technologies. This review underscores the urgent need to involve patients in the design and evaluation of these tools, given the increased access to free tools available to the public to inform future application design, policy development, and safe and effective usage.