This study addresses the normalization of electromyographic signals analyzed during the stance phase of running. The work compares the classical normalization method based on peak muscle activation (Peak%) with two alternative methods, considered innovative: one based on ground reaction force (Frs%) and the other on ground reaction force relative to body weight (Frs%/bw). The muscles evaluated bilaterally were the gluteus medius and the vastus medialis of the quadriceps, in healthy runners, on a treadmill, over a distance of 400 m at a self-selected comfortable speed. The comparison among the three normalization methods was carried out using a statistical model focused on reliability analysis. Reliability was assessed through the calculation of the Intraclass Correlation Coefficient (ICC), reporting both ICC (Random) and ICC (Average Random) for each muscle and method during the stance phase, along with the corresponding p-values. Consistency was evaluated using the Reliability Coefficient (RC). The results indicate that all three methods are reliable and consistent; however, for applications in which consistency and precision are critical, the Frs% and Frs%/bw normalization methods are preferable to the Peak% method.