Abstract
Objectives: Generative artificial intelligence (AI) is increasingly used in higher education, yet evidence regarding its use among physical therapy (PT) students remains limited, particularly in Japan. This study aimed to describe patterns of generative AI use, its applications in PT-related learning and national licensure examination preparation, student perceptions, and institutional guidance across 4 Japanese PT programs. Methods: A cross-sectional survey was conducted among undergraduate PT students from 4 PT educational institutions in Japan (September 2025–February 2026). Of 775 invited students, 525 responded (67.7%); of these, 508 (96.8%) reported prior generative AI use and formed the analytic sample. The remaining 17, who had no experience with generative AI, were excluded from analyses of use patterns, purposes, and perceptions. Comparisons used chi-square tests with Cramér’s V and Mann−Whitney U tests with rank-biserial correlation. Results: ChatGPT (free version; OpenAI, San Francisco, CA, USA) was used by 93.1% of AI users; 61.6% reported weekly use for PT-related studies. Use for national licensure examination preparation increased with academic year (lower years: 57.4% vs. upper years: 81.3%; V = 0.256), with knowledge-confirmation activities predominating. Students rated learning efficiency highly (median = 4 on a 1–5 scale) while also expressing concern about overdependence (median = 4). Only 15.4% were aware of official institutional guidance, with significant variation across institutions (p <0.001). Conclusions: Generative AI use was common among the PT students surveyed. Institutions should develop frameworks promoting its appropriate use as a scaffolding tool for critical thinking, clinical reasoning, and information verification, while establishing clear standards for academic integrity and patient confidentiality.
