Written statements summarizing key facts and arguments can enhance the impacts of mini-publics on policymaking and public discourse. However, the production of such statements is often cumbersome, especially in larger scale mini-publics such as Citizens’ Assemblies. Generative AI tools based on large language models (LLMs) could be helpful in some of the key processes involved in co-writing citizens’ statements, for example, in formulating questions for experts, as well as summarizing, prioritizing and editing arguments to be included in the statement. In the pilot study of Citizens’ Assembly on Energy Consumption, organized in Finland in winter 2025, we test a variety of ways of using generative AI in a deliberative process to support the production of a written statement. Based on participant and facilitator surveys as well as observations at different stages of the deliberative process, we evaluate benefits and pitfalls of AI-assisted co-writing and decision-making in mini-publics and reflect on the optimal division of labor between human deliberators, facilitators and AI systems.
