Abstract We compared the floristic composition and structure of restoration areas of eucalypt woodland with untreated pasture (control) and remnant vegetation (reference) in western Sydney. The restored areas comprised over 1,000 ha of abandoned pasture, which had been treated to reduce weeds and planted with seedlings of 26 native plant species raised from seed obtained locally from remnant vegetation. Plantings were carried out 0–9 years ago. Floristic composition was measured in quadrats using frequency scores and cover abundance. As far as possible treatments and restoration ages were replicated across sites. Ordination and analyses of similarity failed to distinguish the composition of restored vegetation from that of untreated pasture, which were both significantly different from that of remnant vegetation. There was a weak compositional trend with age of restored vegetation, but this was not in the direction of increasing resemblance to remnant vegetation. There was some evidence for convergence in structural features of restored with remnant vegetation, but this was at least partly attributed to plant growth. Subject to constraints imposed by the sampling design, environmental factors, and spatial variation were discounted as explanations for the results. The results therefore suggest either failure of restoration treatments or a restoration trajectory that is too slow to detect within 10 years of establishment. Our conclusions agree with those of similar studies in other ecosystems and support: (1) the need to monitor restoration projects against ecological criteria with rigorous sampling designs and analytical methods, (2) further development of restoration methods, and (3) regulatory approaches that seek to prevent damage to ecosystems rather than those predicated on replacing losses with reconstructed ecosystems.