The Cr, Mn, and Ni content of medium and low alloy steel were analyzed using energy dispersive X-ray fluorescence spectroscopy (EDXRF) and black propagation neural network optimized by genetic algorithm (GA-BP). EDXRF was used to excite the six standard samples of medium and low alloy steel and the X-ray fluorescence spectra were obtained. The characteristic peak intensity of each element was obtained by subtracting the background using the two-point method. A total of 108 groups of spectral data and their corresponding content-based GA-BP neural network were obtained. To forecast the contents of 36 low alloy steel samples, the training completion of the GA-BP neural network was used. The predicted results and the fundamental parameter method analysis results were compared. The average errors of the chemical analysis results of the standard samples were 0.0287%, 0.0314%, and 0.0423% for Cr, Mn, and Ni, respectively. The experimental results showed that the BP neural network optimized by the genetic algorithm is suitable for the EDXRF analysis of Cr, Mn, and Ni in medium and low alloy steel.