Many animal taxa respond strongly to spatial and temporal variation in vegetation structure and floristic composition, suggesting that changes in vegetation could be a cheap and readily observable surrogate for changes in animal assemblages. Yet there is considerable uncertainty about how different taxa respond to vegetation over time, potentially limiting the application of habitat-based surrogates to many areas of applied ecology. We examined the strength and temporal consistency of habitat-based surrogates of three different vertebrate taxa in a landscape with multiple vegetation types. We used regression models to quantify the relationships between six vegetation attributes (species richness and percentage cover of overstorey, midstorey and understorey) and three measures of bird, mammal and reptile assemblages (abundance, species richness and composition). We found that overstorey richness and cover had both consistent and strong positive effects on bird assemblages. Vegetation effects were generally weaker and more variable for mammals and reptiles compared with birds. Each taxon displayed different temporal dynamics following fire, with negative effects on birds and mammals, but positive short-term effects on reptiles. Surprisingly, fire increased vegetation effects on birds, but did not consistently alter vegetation effects on mammals or reptiles, indicating a lack of concordant responses among taxa. Synthesis and applications. Empirical testing of habitat-based surrogates of multiple animal taxa is needed to identify reliable and consistent management proxies. Our study suggests that habitat-based surrogates could be useful metrics for quantifying changes in bird assemblages through time and after fire, but that the same metrics could not be applied to mammal and reptile assemblages. The absence of both strong and consistent effects of vegetation attributes across the three groups of vertebrates suggests that taxon-specific habitat surrogates may be required to detect changes over time and after disturbance within heterogeneous landscapes.