Transformation of Home Physical Therapy Guidance with Generative AI

Patient guidance provided by software tools needs to be based on trustful medical knowledge, to be delivered in a friendly and easy to understand way, and to fit various backgrounds. This paper explores the paradigm shift from traditional decision support tools assisting home based physical therapy toward generative artificial intelligence frameworks, to increase autonomy of the rehabilitation activity. After integrating a GPT-4.1-mini model API into an existing software application, a comparative evaluation is realized between the guidance provided with and without information provided by generative AI, following a physician's diagnosis and an exercise program recommended by a physical therapist. The Flesch-Kincaid index was employed for testing the guidelines complexity and readability, showing how appropriate prompts sent to the API by the application can improve the results.