Real‐time biofeedback gait retraining has been reported to be an effective intervention to lower the impact loading during gait. While many of the previous gait retraining studies have utilized a laboratory‐based setup, some studies used accelerometers affixed at the distal tibia to allow training outside the laboratory environment. However, many commercial sensors for gait modification are shoe‐mounted. Hence, this study sought to compare impact loading parameters measured by shoe‐mounted and tibia sensors in participants before and after a course of walking or running retraining using signal source from the shoe‐mounted sensors. We also compared the correlations between peak positive acceleration measured at shoe (PPAS) and tibia (PPAT) and vertical loading rates, as these loading rates have been related to injury. Twenty‐four and 14 participants underwent a 2‐week visual biofeedback walking and running retraining, respectively. Participants in the walking retraining group experienced lower PPAS following the intervention (P < 0.005). However, they demonstrated no change in PPAT (P = 0.409) nor vertical loading rates (P > 0.098) following the walking retraining. In contrast, participants in the running retraining group experienced a reduction in the PPAT (P = 0.001) and vertical loading rates (P < 0.013) after running retraining. PPAS values were four times that of PPAT for both walking and running suggesting an uncoupling of the shoe with tibia. As such, PPAS was not correlated with vertical loading rates for either walking or running, while significant correlations between PPAT and vertical loading rates were noted. The present study suggests potential limitations of the existing commercial shoe‐mounted sensors.