Harmful algal blooms (HABs) are increasingly affecting the structure and function of temperate marine ecosystems, yet their system-wide impacts and recovery dynamics remain poorly understood. We developed a whole-of-ecosystem food web modelling framework using Ecopath with Ecosim to reconstruct pre-bloom conditions and simulate HAB impacts and recovery trajectories for South Australia’s gulfs. Two mass-balanced Ecopath models were developed for Gulf St Vincent and Spencer Gulf, representing interactions among 98 key trophic groups, including commercially targeted species, habitat-resolved fish assemblages, invertebrates, predators and aquaculture components. HAB impacts were parameterised using a combination of in situ and remotely sensed indicators, fisheries rapid assessments, and fish kill records to derive a spatially and temporally explicit mortality index. Uncertainty in group-level sensitivities was incorporated through an ensemble modelling approach in Ecosim, enabling projections of short- and long-term impacts on trophic structure, network integrity, and recovery timeframes. Scenarios of bloom recurrence, intensity, and potential management interventions were used to explore alternative recovery pathways. Model outputs provide a basis for identifying trade-offs in ecosystem responses under current and future conditions, supporting decision making to optimise ecosystem resilience and guide recovery from the 2025 HAB event.