Equipping the nutrition workforce with the capacity to advocate for evidence-based solutions is essential to improve public health outcomes. The Public Health Advocacy Institute of Western Australia developed a toolkit with key principles of effective advocacy to support the workforce (1). However, opportunities to apply these principles in an authentic environment are rare, typically high-stakes, and viewed as intimidating for inexperienced practitioners. This study examined the effectiveness of a simulation-based advocacy experience using virtual policy-maker avatars on Monash University’s ATLAS platform. Offered to final-year nutrition and dietetics students (n=68), this intervention involved a pre-class online module, a three-hour interactive workshop, and an assessment task where students were required to present their advocacy pitch and respond to questions from their chosen avatar, either a Federal, State or Local government virtual decision-maker. A convergent parallel mixed-methods design evaluated changes across three timepoints: baseline (T1), post-workshop (T2), and post-simulation (T3). Quantitative surveys measured advocacy knowledge, skills, professional identity, and future intentions on a 1–10 scale, analysed using Friedman and Dunn's post-hoc tests. Qualitative analysis of 68 structured reflections identified factors that students described as helping or hindering their learning. Students most commonly chose the Federal avatar (47%, n = 32), followed by State (37%, n = 25) and Local (16%, n = 11). Popular public health ‘asks’ were a tax or levy for sugar sweetened beverages (n=13) and strategies to improve food security (n = 13) such as income support measures, price subsidies for core foods and affordable/accessible food retail options. Significant improvements occurred across the study period for advocacy knowledge (T1 median 6.0 [IQR 5–7] to T3 9.0 [IQR 7.25–10], p < 0.001) and advocacy skills self-efficacy, which showed the largest change (T1 5.0 [IQR 4–6] to T3 8.0 [IQR 7–9], p < 0.0001). Professional identity also increased significantly. The workshop was the primary driver for initial gains, while the ATLAS simulation provided additional perceived value despite ceiling effects. Students rated the simulation highly for practice value and realism. Qualitatively, students reported that the "safe-to-fail" environment emulated real-world advocacy and supported iterative learning through built-in performance feedback. While some students found the unpredictability of political and economic questions challenging, most valued the opportunity to identify "blind spots" in their preparation. Technology comfort gaps present at baseline were eliminated by T3, suggesting the intervention levels the playing field for technology-anxious learners. This research demonstrates that integrating generative AI simulations with structured educational activities can increase advocacy capacity. By providing authentic, low-stakes opportunities, such simulation activities may contribute to a future workforce that is better prepared for the nuances of real-world policy engagement. Further research should determine if these gains in self-reported capability translate into actual long-term advocacy behaviour.