The accuracy of the results of steel x‐ray fluorescence (XRF) analysis using certain computation algorithms, differing in the way in which they consider the dependence of α coefficients on sample composition, was compared by means of the determination of chromium in steel standard samples, the chemical compositions of which represent possible compositional variations of 456 different types of high‐speed, high‐, medium‐ and low‐alloy steels. It is shown that in the XRF analysis of steels, the best accuracy is provided by algorithms in which this dependence is based on real variations in the concentration of interfering elements. Algorithms using constant influence coefficients may also give acceptable accuracy of XRF results provided that more exact calculation of equation parameters, by which the values of reduced element concentrations are obtained, are found. Recommendations for increasing the accuracy of these parameter calculations are given.