Photoluminescence excitation spectroscopy (PLE) can be a very useful in-line metrology tool for photovoltaic manufacturing of III-V and chalcogenide-based, direct gap solar cells. We report on the experiment and numerical modeling of PLE measurements on GaAs-based test structures. Strong suppression of the band edge photoluminescence (PL) intensity is observed in bare, unpassivated GaAs samples, compared to an AlGaAs/GaAs double heterostructures (DH). Similarly, an observed reduction of PL intensity of the unpassivated samples with increasing excitation energy is associated with surface defects. A numerical drift-diffusion model is developed and two frameworks to describe the surface quality are studied, one involving the surface recombination velocity and one including a single trap level. For lightly-doped samples, where the effect of interface trap charging becomes important, the two simulation approaches lead to different spectroscopic response. For the examined samples, an estimated trap density of N-T similar to 10(13)/cm(2) is deduced.