This report describes an exercise to investigate the robustness and reliability of using generative AI to help produce rapid evidence reviews. The AI-assisted output was completed in 23% less time than the human-only output but was judged to be less fluent and required more revisions.