This study presents a global sensitivity analysis (SA) of the near‐fault probabilistic seismic‐hazard analysis (PSHA) model. The influence of the model inputs on the computed PSHA results was investigated using the information‐theoretic SA approach. This process is utilized at three different sites in the city of Tehran, the capital of Iran, to identify the most important uncertain inputs by using an entropy‐based sensitivity index as a measure of importance. The analysis showed that the degree of sensitivity of the input variables depends on the site of interest, return period, and structural periods. Using the results obtained, the ground‐motion prediction equations, the minimum magnitude, and b‐value variables most influence the uncertainty in the hazard estimation. Moreover, the maximum magnitude and the variables that were used to model the near‐fault directivity have a significant contribution at near‐fault sites.