Evaluating ChatGPT’s advice and recommendations regarding exercise for people with inclusion body myositis

The use of Large Language Models for exercise prescription in the general population has been shown to be safe and effective but not comprehensive compared to that of physiotherapists. In rare disease groups, where experts are less accessible, Large Language Models such as ChatGPT may be well-placed to advise on exercise prescription. Inclusion body myositis, a rare, inflammatory myopathy of older people, is unique in both its affected muscle groups and in recommended resistance exercise regimens. This study aimed to determine the accuracy, safety and appropriateness of advice provided by ChatGPT to six questions relating to muscle involvement, resistance exercises and strategies to reduce falls. ChatGPT's responses were reviewed by 12 physiotherapists experienced in inclusion body myositis management. Responses were rated on a five-point rubric ranging from "comprehensive and accurate" to "wrong and unsafe." Responses were judged safe and mostly accurate in five of the six domains. Mostly, ChatGPT was safe and accurate, but at times lacked comprehensiveness. In some areas, however, ChatGPT performed very well, and there is significant potential for its use as a tool to aid exercise prescription in this population.