Generative AI is poised to revolutionize citizen-initiated mechanisms of direct democracy. This article argues that AI functions as a powerful accelerator, lowering historical barriers and making these mechanisms not only cheaper but also more frequent, by automating law-drafting, optimizing mobilization, and enabling hyper-personalized persuasion. However, this efficiency threatens the core conditions of democratic legitimacy. By eroding deliberation, weakening civil society, and corroding trust through synthetic content, AI risks converting citizen-initiated mechanisms of direct democracy from essential “safety valves” into engines of plebiscitarian instability. This essay contrasts a dystopian future of Automated Plebiscites with a preferable path of Augmented Deliberation. To steer toward the latter, the essay proposes a governance roadmap of digital guardrails—including AI watermarking, public-interest AI platforms, and independent algorithmic audits—aimed at ensuring AI augments rather than undermines democratic practice.
