Abstract Purpose Chatbots based on large language models (LLMs), like ChatGPT, have promise to augment care of older people. The evidence base, however, is small. The aim of this systematic review is to study the evidence base of patient-facing LLM chatbots in the care of older people. Methods Following the PRISMA guidelines, a systematic search was conducted in PubMed, Embase, Web of Science, CINAHL, and Cochrane Library up to 1st May 2025 (PROSPERO—CRD42025638985). Studies involving patient-facing LLM chatbots and patients aged 60 years or older were included. Characteristics of the study and chatbots were extracted including their technology readiness levels (TRL). The mixed methods appraisal tool (MMAT) was used to assess the quality of included studies. Results Out of 1228 records, 9 studies were included, with a median sample size of 12 older participants. Of these, 5 were mixed methods, 3 were qualitative, and 1 was a pilot/feasibility study. Seven studies evaluated chatbots providing supportive conversations or social/emotional support. GPT-3.5 (OpenAI) was the most frequently employed LLM. Median TRL was 7; no LLM chatbot was assessed at a TRL of 8 or higher (tested for effectiveness). Conclusions Few studies report on patient-facing LLM chatbots in the care of older people. Most studies utilized qualitative methods. No study evaluated the clinical or cost-effectiveness of chatbots, and there were no studies conducted while the chatbot was fully implemented within clinical workflows. This indicates the research is still in its early exploratory phases. This living review will be updated periodically to follow developments in the field.
Johnson et al. (Fri,) studied this question.