Effects of AI-Assisted Physical Exercise on the Health of Elderly Women: A Randomized Controlled Trial Based on Smart Devices and Personalized Exercise Guidance

Background: Elderly women face significant health challenges, including knee osteoarthritis (KOA) and balance disorders. Artificial intelligence (AI)-assisted exercise intervention can address limitations of traditional intervention methods, such as low compliance and high economic costs. Objective: This randomized controlled trial (RCT) evaluated the effects of AI-assisted Baduanjin exercise on physical health (balance and knee function) in elderly women, comparing it with offline manual guidance and health education. The group of 79 elderly women (60–74 years) were randomly assigned into three groups: AI-assisted Baduanjin (AI group, n = 25), offline instructor-led Baduanjin (Offline group, n = 27), and health education (Education group, n = 27). Methods: Interventions lasted 12 weeks, with three 45-min sessions per week. Two outcome measures were evaluated pre- and post-interventions: postural stability assessed by the unipedal stance test and knee function measured using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). This study considers two measurement methods. One is a two-way repeated-measures analysis of variance used to evaluate the effects on the three intervention groups. The other is an independent-samples t-test, with post hoc testing (Bonferroni), used to assess differences among the three groups. Results: Both the AI and Offline groups showed significant improvements in WOMAC pain and function scores at 12 weeks (p < 0.05), with the Offline group demonstrating greater functional improvement (decrease in WOMAC function score: 6.7 points, Cohen’s d = 1.23, 95% CI 0.81–1.65). No serious adverse events (e.g., falls, exacerbation of joint pain) were reported in any group. The Offline group also showed immediate balance enhancement (closed-eye stance improvement, effect size d ≈1.57), while the AI group exhibited progressive pain relief. The Education group showed minimal improvements. Inter-group comparisons showed the AI and Offline groups outperformed the Education group in balance and knee function (p < 0.05). Conclusions: AI-assisted and offline Baduanjin interventions effectively improve balance and knee function in elderly women, with offline guidance offering improvement of balance ability. AI intervention is suitable for rural elderly women with low digital literacy, as it provides simplified operation and voice prompts to ensure adherence.