Summary In azimuthal AVO (amplitude-variation-with-os et) analysis, it is common to compensate for geometrical spreading by applying empirical amplitude gain corrections that may give erroneous results when the overburden is azimuthally anisotropic. Here, we present a synthetic modeling study, in which we test a more rigorous, moveout-based anisotropic spreading correction (MASC) on PP reections from the bottom of an orthorhombic layer. It should be emphasized that the MASC method operates with the eectiv e moveout parameters derived directly from wide-azimuth traveltimes and does not need any other information about the velocity model. The wide-azimuth, long-oset data were computed by the anisotropic reectivit y method, which generates exact waveelds for a stack of horizontal, azimuthally anisotropic layers. Although the MASC algorithm is valid within the framework of ray theory and does not account for the transmission through the top of the target layer, it accurately reconstructs the reection coecien t for a wide range of osets and azimuths. Errors in the estimated reection coecien t are mostly caused by the intereference of the target event with shear and mode-converted waves. We also examined the performance of the empirical t 2 gain correction on two models with dieren t relative strength of the azimuthal variation of the geometrical spreading. If the azimuthal dependence of the reection coecien t is much more pronounced than that of the geometrical spreading (i.e., three times or more at an angle of 20 , as in our model 1), the t 2 -correction can be used for purposes of qualitative azimuthal AVO analysis. However, when this condition is not satised (our model 2), the empirical correction distorts or even obliterates the azimuthal AVO signature.