The study presented in this paper addresses the issue of engineering validation of Graves and Pitarka's (2010) hybrid broadband ground motion simulation methodology with respect to some well-recorded historical events and considering the response of multiple degrees of freedom (MDoF) systems. Herein, validation encompasses detailed assessment of how similar is, for a given event, the seismic response due to comparable hybrid broadband simulated records and real records. In the first part of this study, in order to investigate the dynamic response of a wide range of buildings, MDoF structures are modeled as elastic continuum systems consisting of a combination of a flexural cantilever beam coupled with a shear cantilever beam. A number of such continuum systems are selected including the following: (1) 16 oscillation periods between 0.1 and 6 s; (2) three shear to flexural deformation ratios to represent respectively shear-wall structures, dual systems, and moment-resisting frames; and (3) two stiffness distributions along the height of the systems, that is, uniform and linear. Demand spectra in terms of generalized maximum interstory drift ratio (IDR) and peak floor acceleration (PFA) are derived using simulations and actual recordings for four historical earthquakes, namely, the 1979 Mw 6.5 Imperial Valley earthquake, 1989 Mw 6.8 Loma Prieta earthquake, 1992 Mw 7.2 Landers earthquake, and 1994 Mw 6.7 Northridge earthquake. In the second part, for two nonlinear case study structures, the IDR and PFA distributions over the height and their statistics, are obtained and compared for both recorded and simulated time histories. These structures are steel moment frames designed for high seismic hazard, 20-story high-rise and 6-story low-rise buildings. The results from this study highlight the similarities and differences between simulated and real records in terms of median and intra-event standard deviation of logs of seismic demands for MDoF building systems. This general agreement, in a broad range of moderate and long periods, may provide confidence in the use of the simulation methodology for engineering applications, whereas the discrepancies, statistically significant only at short periods, may help in addressing improvements in generation of synthetic records. Copyright © 2013 John Wiley & Sons, Ltd.