The trends and variance in amplitude variation with offset (AVO) observations caused by near‐surface structure, attenuation, and scattering are numerically synthesized by pseudospectral viscoelastic 2-D modeling. Near‐surface structure produces amplitude focusing and defocusing that significantly distort AVO observations in offset windows at a scale comparable to that of the lateral variations in the structure. Attenuation and scattering decrease absolute amplitudes at all offsets. Scattering and wave interference increase the variance associated with AVO measurements. Depending on the relative influence of intrinsic attenuation, apparent attenuation associated with scattering, and geometrical focusing, a normalized AVO response can increase or decrease with offset (relative to that for the associated elastic, nonscattering, 1-D solution), and so mimic the behavior predicted as a function of contrasts in density, velocity, porosity or Poisson’s ratio. If only relative (normalized) amplitudes are available, it is difficult to distinguish between effects of parameters whose main contributions are to absolute amplitude; for example, a trend of decreasing amplitude (relative to that for an elastic flat‐layered model) produced by intrinsic attenuation may be counteracted by focusing/scattering or anisotropic effects over wide aperture ranges. Diagnostic information on AVO effects of scattering and attenuation is lost when the noise level is sufficiently high. Interpretations of AVO observations based on homogeneous layered elastic models must therefore be used with caution as they are, in general, nonunique. Lateral variations in AVO parameters are the key to detecting hydrocarbons, so lateral changes in AVO produced by lateral changes in the overburden properties have potential for being misinterpreted, especially if the recording aperture is small.