Although previous studies have performed finite‐fault simulations of actual or hypothetical earthquakes to generate time histories of near‐fault ground strains and rotations, no systematic attempt has been made to assess the sensitivity of these motions to variations in seismic source parameters (e.g., fault type, magnitude, rupture velocity, slip velocity, hypocenter location, burial depth). Such a parametric investigation is presented in this article by generating time histories of ground strains and rotations at near‐fault stations and at a dense grid of observation points extending over the causative fault for a suite of hypothetical strike‐slip and dip‐slip earthquakes. The simulation results show that strike‐slip earthquakes produce large shear strain and torsion, whereas dip‐slip earthquakes generate large axial strain and rocking. The time histories of specific components of displacement gradient, strain, and rotation at near‐fault stations may be estimated from those of ground velocities using a simple scaling relation, whereas peak rotational motions in the near‐fault region may be reasonably estimated from peak translational motions using a properly selected scaling factor. The parametric analysis results show that near‐fault ground strains and rotations exhibit strong sensitivity to variations in rupture velocity, slip velocity, and burial depth, whereas a change in hypocenter location significantly alters the spatial distributions of peak ground strains (PGSs) and rotations (PGRs). The presence of a low‐velocity surface layer increases the amplitude and duration of ground strains and rotations, whereas their static offsets are also amplified. Distinct attenuation characteristics are observed for PGSs and PGRs depending on the component of interest, the earthquake magnitude, and the rupture distance. Finally, the spatial distributions of PGSs and PGRs obtained from a stochastically generated variable slip distribution are overall similar to those obtained from a tapered uniform slip distribution, whereas the spatial distributions of the respective static offsets differ significantly.