Background Diaphragmatic dysfunction contributes to exercise intolerance in COPD, but whether BMI shapes baseline function and response to exercise remains unclear. This secondary analysis characterized BMI-related diaphragmatic phenotypes and whether diaphragm thickening fraction (TFdi) links exercise training to functional improvement. Methods Data were analyzed from 103 patients with stable COPD in a completed 2×2 factorial randomized trial. Because no exercise-by-acupuncture interaction was detected for 4-week TFdi, participants were regrouped as exercise-exposed or non-exercise. Associations between BMI and baseline TFdi were assessed using restricted cubic splines and threshold models. Regression models estimated exercise effects on 4-week TFdi and 6-minute walk distance (6MWD), with BMI-stratified interaction and exploratory mediation analyses. Results BMI showed a nonlinear association with baseline TFdi, with an adjusted turning point at 23.7 kg/m². Below this threshold, BMI was positively associated with TFdi (β=0.05, 95% CI 0.03–0.08); above it, the association was negative (β=−0.05, 95% CI −0.07 to −0.03). Exercise was associated with higher 4-week TFdi (β=0.15, 95% CI 0.01–0.29) and longer 6MWD (β=59.61 m, 95% CI 10.65–108.57). The TFdi response was most pronounced at BMI ≥25.0 kg/m², with borderline interaction evidence. Exploratory mediation suggested TFdi partially mediated exercise-related 6MWD gains, particularly in this BMI group. Conclusions This study integrates BMI phenotyping, diaphragm ultrasound, and mediation analysis to show that diaphragmatic function and exercise-induced diaphragmatic adaptation in COPD are BMI dependent. BMI may help identify patients most likely to derive diaphragmatic benefit from rehabilitation.