A 0.05° × 0.05° gridded dataset of daily observed rainfall is compared with high-quality station data at 119 sites across Australia for performance in capturing extreme rainfall characteristics. A range of statistics was calculated and analysed for a selection of extreme indices representing the frequency and intensity of heavy rainfall events, and their contribution to total rainfall. As is often found for interpolated data, we show that the gridded dataset tends to underestimate the intensity of extreme heavy rainfall events and the contribution of these events to total annual rainfall as well as overestimating the frequency and intensity of very low rainfall events. The interpolated dataset captures the interannual variability in extreme indices. The spatial extent of significant trends in the frequency of extreme rainfall events is also reproduced to some degree. An investigation into the performance of this gridded dataset in remote areas reveals issues, such as the appearance of spurious trends, when stations come in and out of use. We recommend masking over areas of low station density for this particular gridded data. It is likely that in areas of low station density, gridded datasets will, in general, not perform as well. Therefore, caution should be exercised when examining trends and variability in these regions. We conclude that this gridded product is suitable for use in studies on trends and variability in rainfall extremes across much of Australia. The methodology employed in this study, to examine extreme rainfall over Australia in a gridded dataset, may be applied to other areas of the world. While our study indicates that, in general, gridded datasets can be used to investigate extreme rainfall trends and variability, the data should first be subjected to tests similar to those employed here.