Biodiversity offsets are a popular policy tool for mitigating the impact of development on biodiversity, but the ecological success of offsets arise from complex interactions among socio-economic, ecological and policy processes, making outcomes challenging to assess. Many offset policies use habitat surrogates to determine offset requirements, rather than using direct measures of impacted biota, and this can lead to poor outcomes for species. One potential solution to this is for offsets to be delivered by a public agency (agency-led) rather than by developers (developer-led). This is because agencies may be able to strategically choose offset sites that maximise outcomes for species (e.g. abundance), while there may be little reason for developers to act strategically in this way when offset requirements are based purely on habitat surrogates. Yet, the success of a strategic agency-led approach is likely to depend on patterns of development and offset site availability. To examine this, we developed a novel integrated spatially explicit model of land-use change, habitat, species abundance and offset regulation. We apply the model to the Queensland Government's Environmental Offsets Policy for koalas Phascolarctos cinereus in South East Queensland, Australia, and test how patterns of development and offset site availability influence the performance of agency-led versus developer-led offsets. When potential offset sites were plentiful, agency-led offsets tended to outperform developer-led offset delivery for maximising koala abundance while achieving similar or better outcomes for habitat area. Yet, when potential offset sites were rare, the relative performance of agency-led offset was often poor, and offset requirements for habitat area were less likely to be met. Different spatial patterns of development had little effect on the relative performance of agency-led versus developer-led offsets. Our analysis shows that agency-led offsets with strategic choices of offset sites can improve species' outcomes for habitat-based offsets but can also risk failing to meet habitat area requirements when the availability of offset sites is low. Importantly, our integrated spatial model provides a holistic approach to assessing policy options for biodiversity offsets in dynamic human-modified landscapes. Read the free Plain Language Summary for this article on the Journal blog.