The horizontal‐to‐vertical spectral ratios from microtremor (mHVSR) data obtained at 196 seismic stations in California are used to evaluate three alternative microtremor‐based proxies for site amplification for use in ground‐motion models (GMMs): the site fundamental period (⁠f_0⁠), the period‐dependent amplitude of the mHVSR(T), and the normalized amplitude of the mHVSR(T). The alternative parameters are evaluated for the sites with and without measurements of V_S30⁠. If a V_S30 measurement is not available for a site, then f₀ has the highest correlation with the site amplification for short periods (T <1 s) and the normalized amplitude of the mHVSR(T) has the highest correlation for long periods (T ≥1 s). If a measurement of the V_S30 is available for a site, then the normalized amplitude of the mHVSR(T) has the highest correlation for the site amplification not explained by V_S30 for all periods. For both cases, the correlations are strongest at the longer periods as mHVSR(T) measurements excel at providing valuable information for sites with long‐period amplification due to the deeper velocity structure. In particular, for sites with a V_S30 measurement, the normalized mHVSR(T) amplitude provides more information about the long‐period site terms than the basin depth currently used in GMMs. Empirical models of the median and standard deviation of the site terms based on the normalized mHVSR(T) curves are developed for the two cases. These models can be used directly in the ASK14 GMM to modify the median and aleatory standard deviation or they can be used to estimate the site‐specific site term in the context of a partially nonergodic GMM. Including the mHVSR(T) measurement can have a significant effect on estimates of the ground motion at a site: the range 5%–95% on the observed HVSR(T) values corresponds to factors of 0.6–1.6 for the median spectral acceleration for periods between 0.5 and 4 s.