Democracy is facing challenges including loss of trust in EU institutions, disillusionment and declining interest among citizens, increasing political polarisation, online disinformation and politically manipulated information, and a growing distance between citizens and elected representatives. We believe this can change if proper deliberative processes and tools are institutionalised that exploit the rapid advances in Artificial Intelligence and citizens’ post-Covid competences. We envision the next generation of digital deliberations as: grounded on solid democracy theory, multimodal (video, audio, text), usable and accessible, gamified (e.g. using skill points, scoreboards, missions, and avatars), mass, enabled by novel but also legal and ethical AI features (e.g., summarisation, moderation, fact checking, hate speech and toxicity detection) and easily adopted and institutionalized. In this context, the aim of AI4Deliberation project is to provide robust, ethical AI tools and comprehensive guidance to assist governments in institutionalising, using and evaluating multimodal, gamified, mass deliberations. The vision of the project is to equip governments with a theoretically solid and empirically tested set of AI-enabled deliberative processes, a comprehensive framework with practical guidelines, and an AI toolkit that will enable them to design, institutionalise, operate and evaluate transparent, ethical, inclusive, multimodal, gamified, mass citizens deliberations resulting in more active and inclusive citizenship and increased trust to rule-of-law based institutions by citizens. To achieve these ambitious objectives, the consortium brings together a team of world leaders in deliberative democracy, AI and LLMs (incl. vid-LLMs), argumentation mining, law and ethics, deliberation platforms (including video-based) while four large scale pilots will be conducted from city level to international discussing topics that include climate change and long Covid.