We examine the applicability of the Next Generation Attenuation‐West2 (NGA‐West2) site‐effects model (Seyhan and Stewart, 2014), which is a VS30‐based (time‐averaged shear‐wave velocity in the upper 30 m) model, to sites in central and eastern North America (CENA), using the NGA‐East ground‐motion database. We determine residual site terms by comparing the observed CENA ground‐motion amplitudes, adjusted to B/C site condition (VS30=760 m/s) using the western site‐effects model, to the corresponding predicted amplitudes of a CENA ground‐motion prediction equation for B/C site condition (Yenier and Atkinson, 2015). The CENA prediction model used the same western site model to level the database before model development. Thus, residual trends reveal inadequacies of the western site‐effects model when applied to CENA ground motions. Plotting the residual site terms versus their corresponding site fundamental frequencies (fpeak) reveals significant fpeak‐dependent trends at all frequencies. Average residual site terms for CENA sites, after the western site amplifications have been removed, can be as large as 0.45 in log10 units (2.8 in nonlog units) around f≈fpeak. Correlating the site terms with western site‐effects predictions reveals that at f<1 Hz, site terms in CENA scale with VS30 in a manner that is similar to the way they behave in the NGA‐West2 database. However, at higher frequencies, the correlation of site amplification with VS30 decreases markedly in CENA. In contrast, the horizontal‐to‐vertical (H/V) component spectral ratio (as obtained from the NGA‐East ground‐motion data) is a good predictor of the observed fpeak‐dependent site terms, suggesting that H/V spectral ratio is a more reliable predictor of site amplification for the recording stations of the NGA‐East database. The results of this study demonstrate the importance of a well‐calibrated regional site‐effects model for CENA and the importance of fpeak as a site indicator.