An assessment of whether rehabilitated mine sites have resulted in natural or novel ecosystems requires monitoring over considerable periods of time or the use of space-for-time substitution (chronosequence) approaches. To provide an assessment of ecosystem recovery in areas mined for bauxite in 1975, the ant fauna of one area planted with Eucalyptus resinifera, one seeded with mixed native species, one topsoiled but unrestored, and a forest reference was subjected to a ‘long-term’ study by sampling monthly and latterly annually between 1976 and 1989 using pitfall traps. These plots were resampled in 2012. A companion ‘short-term’ chronosequence study was performed in 1979 in 28 bauxite mines of various ages and restored by a range of different methods, plus three forest references. In order to examine the assertion that the observed differences between restored areas and forest references will lessen with time, sampling using comparable methods was repeated in 2012 in seven of the original plots, representing progressive advances in rehabilitation technology: planted pines; planted eastern states eucalypts; planted native eucalypts; planted eucalypts over seeded understorey; and planted eucalypts on fresh, double-stripped topsoil, plus two forest reference sites. Ant and other invertebrate richness in the long-term study was initially superior in the seeded plot, with little difference between the planted and unplanted plots. It was concluded that although composition of the ant fauna had converged on that of the forest over the 14-year period, differences still persisted. The 2012 resampling revealed that ant species richness and composition had deteriorated in the seeded plot, while values in the unplanted plot, which now supported naturally colonised trees and an understorey, had increased. Differences between all rehabilitated plots and forest still persisted. As with the long-term study, the rate of fauna return and the type of ants present in the short-term study plots differed with the method of rehabilitation used, and, in 1979, no plots had converged on the forest in terms of the ant assemblage. By 2012 ant richness increased, and more so with each advance in rehabilitation technology, except for seeding, in which the understorey had collapsed. Double-stripping of topsoil resulted in the greatest improvements in ant species richness, although none of the areas had converged on the forest reference areas in terms of assemblage composition or ant functional group profiles. Furthermore, assemblage composition in the forest had changed over time, possibly due to reductions in rainfall, which further complicates rehabilitation objectives. It is concluded that although rehabilitation can achieve its objective of restoring diversity, the original assemblage has still not been achieved after 37 years, suggesting that a degree of novelty has been introduced into these older-style rehabilitated areas. The company’s current rehabilitation practices reflect multiple advances in their approach, lending optimism that current restoration may achieve something close to the original ecosystem, an outcome that can only be verified by extended studies like the one described here.