Evaluating the daily precipitation performance of general circulation models at a regional scale is beset with two (related) difficulties: a) ensuring an adequate network of observed daily station data across a region, and b) establishing 'true' areal means of daily precipitation from these station data to ensure a like-with-like comparison. These two difficulties have generally restricted past evaluations of daily model precipitation to individual grid boxes and to comparisons of model grid-box data with observed sub-grid-box scale data. In this paper we exploit a dataset of several dozen daily station time series in western Europe using a new method for estimating 'true' areal mean daily precipitation to evaluate the daily precipitation performance in the region of 12 atmospheric general circulation model simulations undertaken as part of the Atmospheric Model Intercomparison Project. We examine four quantities derived from the simulated daily precipitation: seasonal mean precipitation, daily standard deviation, the frequency of raindays and the mean precipitation intensity of those raindays. We show that simulated winter precipitation tends to be too great, falls on too many days, but is generally less intense than that observed. In summer, model results are more variable with some model simulations yielding statistics quite similar to those observed. Although there is some uncertainty in these evaluation statistics owing to the effect of inter-decadal climate variability, there is a weak suggestion of improved performance with increased spatial resolution. Any systematic effect in model performance due to different convective parameterisation schemes is obscured by other confounding factors.