<p>The article examines how the widespread societal adoption of AI applications is transforming democratic participation and representation. Unlike dominant debates on the manipulation of the political public sphere or the automation of state action, the article focuses on the reconfiguration of democratic participation processes and the institutional processing and feedback of political inputs. To this end, the article proposes a&#xa0;heuristic model that categorises key forms of use along three dimensions: the direction of control, the logic of interaction, and the institutional scope. Four pathways are distinguished: top-down interfaces that scale institutional interaction with members of the public; throughput support in the context of democratic innovations; bottom-up applications that expand civil society mobilisation; and system-replacement models that delegate participatory functions to personal agents. For each of these pathways, the article discusses the resulting research questions, as well as design and governance challenges.</p>