The spread of buffel grass (Cenchrus ciliaris) in semi-arid Australia in recent decades has substantially increased ground cover and fuel loads, particularly in open woodland vegetation communities. The resulting alteration of fire regimes may be the most significant impact of buffel invasion on ecological communities in these areas. Broad scale management of buffel grass is currently not an option in Australia but it is becoming increasingly relevant to assess the benefits of restoring areas of native vegetation where preventing buffel grass invasion is no-longer possible. We managed buffel grass in a series of experimental plots from 2008–2012. In June and August 2011, two unplanned fires burnt through the plots providing a unique opportunity to compare the outcome of wildfire, including the spatial pattern of fire, and the effect on ground vegetation and on a long-lived, perennial overstorey species, in replicated managed and unmanaged plots. The area of ground that remained unburnt was much greater in managed plots (with predominantly native vegetation) than unmanaged (predominantly buffel grass) plots and where the managed plots did burn the fire was more patchy. This had direct implications for the richness of ground layer plant taxa following fire and the extent to which overstorey trees were exposed to fire. Fire increased pre-existing differences in the number of taxa in the ground level vegetation, an effect that persisted for the duration of our study, suggesting that fire accelerates direct negative competitive effects between buffel grass and native grasses and forbs. Hakea divaricata (fork-leafed corkwood) trees in unmanaged buffel grass sites suffered higher burn intensities, and their long-term viability at this location is likely to be threatened if fires fuelled by buffel grass continue. Our results demonstrate clear benefits of removing fire-enhancing invasive plants from areas of high conservation value.