Seismic demand estimation of structures (which is considered as one of the main components of performance-based designing) is along with various uncertainties, the most important of which is fault-to-site distance. In fact, with the variation in fault-to-site distance of an accelerogram, its effect on seismic demand estimation would differ. However, it seems that this distance impact on the seismic demand will be different in far and near fields concerning the distinct nature of near field. The main aim of this study is to verify this issue and determine the impact of fault-to-site distance on seismic demand of steel moment resisting frames in near and far fields using nonlinear dynamic analysis and Bayesian statistics. Nonlinear dynamic analysis is used to cover the actual nonlinear behavior of the structure in near-collapse performance level and the Bayesian approach to cover all uncertainties. Concerning the research objectives, two generic steel moment resisting frames of 3-storey with rigid behavior and 15-storey with flexible behavior have been selected and nonlinearly modeled in Open Sees. At the next stage, these frames were analyzed through incremental nonlinear dynamic analysis under five groups of 40 accelerograms that were similar in terms of all features except for fault-to-site distance and the results were used for determining their seismic demand. In so far as the only variable in this analysis is fault-to-site distance, the difference in the results could be attributed to this variable. According to the results, from statistical approach, there is some difference between the impact of distance variation on seismic demand in near and far fields which is subject to some variables such as the behavior of the frame and its performance level.