Comparison of the effects of artificial intelligence, virtual reality, and traditional balance training programs on balance and gait in older adults.
Key claim: An 8-week ChatGPT-4o-generated exercise program and a VR exergaming program both produced significantly greater gait speed gains than traditional balance training in older adults.
Abstract
PURPOSE: With ageing, the ability to maintain balance and gait declines, leading to falls and reduced independence. This trial compared an AI-generated, human-delivered exercise program (ChatGPT-4o), a virtual reality (VR) exergaming program, and a traditional balance (TB) training program on balance and gait in older adults residing in retirement homes. MATERIALS AND METHODS: In this stratified, parallel-group, randomised controlled trial, 48 older adults aged 60-75 years were allocated 1:1:1 to VR (Xbox Kinect; = 16), AI (ChatGPT-4o-generated program; = 16), or TB ( = 16). All groups completed 60-min sessions, three times per week for 8 weeks, with a 2-month follow-up. Balance, mobility, balance confidence, gait speed, cadence, Gait Stability Ratio (GSR), and usability (System Usability Scale, SUS) were assessed at all three time points. RESULTS: All groups improved significantly over time (< .001); group × time interactions were significant for all outcomes (ηp= .195-.631) except ABC ( = .063 after Greenhouse-Geisser correction). VR and AI produced greater gait-speed gains than TB (both< .001, no VR-AI difference); VR produced the largest cadence gains and AI the largest GSR reduction. Usability was excellent for VR (SUS 82.4) and good for the instructor-delivered AI program (74.6). No adverse events occurred. CONCLUSION: AI-, VR-, and traditional balance training can each improve balance and gait in older adults, while technology-assisted programs provide additional gait benefits, supporting more personalised, hybrid rehabilitation approaches. CLINICAL TRIAL REGISTRATION: IRCT20181113041645N2 (May 18 2025).
Why this matters
This early randomized trial adds human outcome evidence to AI-designed exercise: an 8-week ChatGPT-4o-generated program and VR exergaming each improved gait speed more than traditional balance training in older adults [human-rct]. The evidence is narrower than an AI-for-longevity claim, leaving durability, falls, and broader aging outcomes unresolved; the supplied summary also lacks effect sizes and adverse-event details [human-rct]. Trend trajectory remains separate: this trial supplies no media or community evidence of consumer adoption.