This study investigates how current practices of input ground-motion selection influence site response analysis results and their variability, when considering different tectonic settings. Study sites in Seattle and Boston are chosen to represent tectonic settings with contributions to the seismic hazard from shallow crustal and subduction events, as well as stable continental regions, respectively. Selected input ground-motion suites for one-dimensional site response analysis represent variations in the target spectrum definition, spectral period of interest, seismic source, and ground-motion database. When directly incorporating different types of seismic sources (e.g. shallow crustal versus subduction) into target spectrum definitions and selecting ground motions from the corresponding databases (i.e. consistent with such seismic sources), differences on the estimated site response and its variability are observed. These effects are captured by spectral amplification factors and nonspectral intensity measures (significant duration and Arias intensity) and become particularly apparent for subduction zones. The variability in spectral amplification factors stemming from ground-motion selection techniques is found to be also a function of the characteristics of the site, becoming higher near the fundamental period of the site. Estimated responses at stiffer sites are more significantly influenced by ground-motion selection techniques, whereas the onset of nonlinear soil behavior at softer sites can reduce such variability.