ChatGPT and DeepSeek-Assisted Rehabilitation in Subacromial Pain

This completed randomized controlled trial investigated the clinical effects of integrating AI-generated exercise recommendations into conventional physiotherapy for individuals with subacromial pain syndrome (SAPS). A total of 82 individuals were assessed for eligibility; 7 were excluded for not meeting inclusion criteria. Seventy-five participants provided consent and were randomized in a 1:1:1 ratio to 3 groups (n=25 each): (1) conventional physiotherapy plus a conventional exercise program, (2) conventional physiotherapy plus ChatGPT-guided exercises, and (3) conventional physiotherapy plus DeepSeek-guided exercises. The allocation sequence was generated using simple computer-based randomization by an independent staff member and was revealed after completion of baseline assessments. Interventions were delivered over 6 weeks (3 supervised sessions/week). Six participants discontinued the intervention due to scheduling conflicts (2 per group). Sixty-nine participants completed the 6-week intervention and were included in the final analysis (n=23 per group). The primary outcome was shoulder pain and disability (Shoulder Pain and Disability Index, SPADI), assessed at baseline and 6 weeks. Secondary outcomes included pain intensity (Numeric Pain Rating Scale, NPRS), health-related quality of life (Short Form-12, SF-12), upper limb function (Disabilities of the Arm, Shoulder and Hand, Quick DASH), range of motion, isometric strength, kinesiophobia, catastrophizing, emotional status, and disease-specific quality of life.