Biodiversity offset schemes are globally popular policy tools for balancing the competing demands of conservation and development. Trading currencies for losses and gains in biodiversity value at development and credit sites are usually based on several vegetation attributes combined to yield a simple score (multimetric), but the score is rarely validated prior to implementation. Inaccurate biodiversity trading currencies are likely to accelerate global biodiversity loss through unrepresentative trades of losses and gains. We tested a model vegetation multimetric (i.e., vegetation structural and compositional attributes) typical of offset trading currencies to determine whether it represented measurable components of compositional and functional biodiversity. Study sites were located in remnant patches of a critically endangered ecological community in western Sydney, Australia, an area representative of global conflicts between conservation and expanding urban development. We sampled ant fauna composition with pitfall traps and enumerated removal by ants of native plant seeds from artificial seed containers (seed depots). Ants are an excellent model taxon because they are strongly associated with habitat complexity, respond rapidly to environmental change, and are functionally important at many trophic levels. The vegetation multimetric did not predict differences in ant community composition or seed removal, despite underlying assumptions that biodiversity trading currencies used in offset schemes represent all components of a site's biodiversity value. This suggests that vegetation multimetrics are inadequate surrogates for total biodiversity value. These findings highlight the urgent need to refine existing offsetting multimetrics to ensure they meet underlying assumptions of surrogacy. Despite the best intentions, offset schemes will never achieve their goal of no net loss of biodiversity values if trades are based on metrics unrepresentative of total biodiversity.