Buffer strips with no added fertilizers are a mitigation measure to reduce P loads from agricultural land to surface water. However, the experimental evidence on their effectiveness (BSE) has been from sloping locations with shallow flow and surface runoff. The aim of this experimental study was to quantify BSE for P on flat agricultural lowland, predominantly with deep flow. We selected sites characteristic of five major hydrogeological classes in the Netherlands and in each experimental field installed paired 5-m-wide unfertilized grass (BS) and reference treatments (REF) that abutted the ditch. The REF treatments were managed similar to the adjacent grass or maize field, but BS treatments were only harvested. Each treatment had a reservoir in the ditch to collect and measure discharge and flow-proportional P concentration for 3 or 4 yr. We also measured net P withdrawal, the P status of the soil and P concentration in upper groundwater. We found a significant BSE for P of 61% on the site with the shallowest flow and steepest slope (2%). At the other sites, BSE was low and statistically insignificant. We conclude that harvested unfertilized buffer strips reduce P loads from flat fields only in specific areas with high surface runoff and/or shallow flow, especially in combination with a high original soil P status.