This paper presents a new empirical model to predict the mean period (T_m) as a frequency-content parameter of earthquake record using the strong ground motions recorded in Iran during 1975–2019. An updated earthquake databank containing 2281 horizontal acceleration records was employed to develop the empirical model through a systematic fitting procedure. A simple functional form for the model was found as a function of epicentral distance (R), moment magnitude (M_w), and the shear wave velocity averaged at the top 30 m of the recording sites (V_s30). The proposed model is compared with three existing predictive models and the results are discussed in terms of magnitude, source-to-site, and site dependencies.