Abstract This paper presents results of a study aimed at evaluating the effect of aftershocks in steel framed buildings. For that purpose, three frame models representing existing steel moment-resisting frames were subjected to a set of as-recorded mainshock–aftershock seismic sequences. For this purpose, 64 as-recorded seismic sequences registered as a consequence of the 1994 Northridge and 1980 Mammoth Lakes earthquakes were considered in this study. In particular, this investigation employed 14 seismic sequences recorded in 7 accelerographic stations in the near-fault region. An examination of the as-recorded seismic sequences shows that the frequency content of the mainshock and the main aftershock is weakly correlated. The response of the frame models was measured in terms of the peak and residual (permanent) drift demands at the end of the earthquake’s excitation. From the results of this investigation, unlike previous results based on artificial seismic sequences, it was found that as-recorded aftershocks do not significantly increase peak and residual drift demands since the predominant period of the aftershocks (i.e. frequency content) is very different from the period of vibration of the frame models. In addition, it was shown that artificial seismic sequences could significantly overestimate median peak and residual drift demands as well as the record-to-record variability.