The aim of this study is to evaluate the velocity semblance analysis technique for air-coupled common midpoint (CMP) measurements with a small antenna offset. The technique was originally developed for seismic surveys, assuming the small spread approximation. Owing to the strong refraction at the surface and shallow investigation depth, this assumption is not valid in the case of air-coupled ground-penetrating radar (GPR). To overcome this assumption, a modification to the method is proposed accounting for the refraction at the surface. Synthetic experiments were executed to demonstrate that the traditional method resulted in a persistent overestimation, whereas the modified method improved the results significantly. Two field experiments have been conducted to test the method under different field conditions. In a first experiment on a road test site, the modified method improved the estimation of depth and propagation velocity significantly. However, the technique failed to estimate the propagation velocity and depth or objects in a second field test, due to rough terrain conditions and noise in the data. Therefore, an additional modification was proposed, by incorporating in-line data as well. This improved the depth and velocity estimations significantly. Overall, this study demonstrates that the traditional velocity semblance analysis (TRAD) is not valid for air-coupled GPR. By accounting for the refraction at the surface and incorporating inline data, it is possible to successfully estimate depth and propagation velocity with small offset air-coupled GPR configurations.