Dynamically downscaled simulations are an attractive source of information for those requiring greater regional detail than what Global Climate Models (GCMs) can provide. However, an important question is whether the information derived from downscaled simulations is plausible. Some recent studies have attempted to answer this question using an ‘added value’ framework. This study takes an alternate approach by comparing projected changes of climate extremes from Regionally Dynamically Downscaled Model (RDM) ensembles to the host GCM ensemble over the Australian continent. Bias-corrected sea-surface temperatures (SSTs) from five Coupled Model Inter-comparison Program Phase 3 (CMIP3) GCMs are used to force two different configurations of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Conformal Cubic Atmosphere Model (CCAM), run at 60 km. For each member of the two CCAM ensembles and each GCM ensemble member, a selection of extreme events and their changes were calculated for minimum temperature, maximum temperature and precipitation for the climate of the 20th century (20c3m; 1980–1999) and A2 (2081–2100) simulations. We find that CCAM shows finer spatial detail in the current distribution of extremes and also the projected climate change signal than GCMs. While the projected change in extreme temperatures has a similar pattern in CCAM compared to the host GCMs, the projected changes in rainfall are not only more spatially detailed but also different at the large scale. This, however, does not infer that climate change signal in CCAM is necessarily more or less physically plausible than the GCMs. In order to ascertain the plausibility and usefulness of RDM simulations, we strongly encourage a critical assessment of the required simulations before their use in adaptation and impacts planning, which includes a comparison to similar simulations from the host GCM(s).