Typical ground‐motion prediction equations (GMPEs) do not account for systematic path effects, which leads to an overestimation of aleatory variability. We provide an update to the Californian GMPE model of Abrahamson et al. (2014) that explicitly accounts for path effects in California by replacing the anelastic attenuation coefficient with a regionally varying one. The updated attenuation model is based on the approach of Dawood and Rodriguez‐Marek (2013). Accounting for path effects leads to a smaller value of the aleatory variability, and results in different median predictions, depending on source and site location. The model is cast as a Bayesian hierarchical model, which allows one to capture the epistemic uncertainties in the anelastic attenuation coefficients. We show that it is important to account for the uncertainty in the attenuation coefficients when using the model for prediction. The differences in attenuation decrease with increasing spectral period. The updated attenuation model is developed with respect to the GMPE of Abrahamson et al. (2014) but can be easily extended for other models and other regions as well.