The newest daily and monthly Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depths (AOD or τ) dataset over land, C005, retrieved using the second-generation operational algorithm, were evaluated using a ground-based Aerosol Robotic Network (AERONET) dataset from 13 sites over China. The dataset covers the period 2003–2006. Daily MODIS C005 AODs over China were found to have a positive bias with a relationship of τMODIS = 0.135 + 1.022τAERONET, for which the offset is larger than reported global validation results. However, the relationship τMODIS = 0.021 + 0.929τAERONET showed that monthly MODIS C005 AODs were an overestimation for small AOD and underestimation for high AOD. Both daily and monthly MODIS AOD retrievals showed poor performance in extreme aerosol conditions, e.g. under dust events or heavy urban/industrial haze. Nevertheless, both daily and monthly MODIS C005 AOD datasets can be used for investigation of aerosol spatial distribution and temporal variation over China.