LLM-driven personalized and adaptive systems for health and wellness: a systematic review

Large language models (LLMs) are increasingly explored as AI systems for building personalized and adaptive health applications. While there has been progress in tailoring LLMs to support diverse populations, inclusivity and contextual sensitivity remain underexplored. In this comprehensive systematic review, we synthesized 149 peer-reviewed articles to examine how LLMs meet individual needs and preferences across diverse health domains. We identified three core dimensions of personalization, eight model adaptation architectures employed, and four categories of implementation challenges that reduce the effectiveness of LLMs. We highlighted the effectiveness of LLM-driven applications across health domains and provided an ethical perspective and opportunities for global impact. We contribute to the broader human-computer interaction (HCI) community by offering a novel conceptual framework to guide the design and development of future LLM-driven applications. The aim is to make these models more adaptive, personalized, inclusive, effective, and responsive to evolving individual characteristics and needs.