Who Governs the Interface? Platform Design, Human–AI Interaction, and Democratic Authority in Civic Technologies

Artificial intelligence is increasingly embedded in civic technologies that mediate communication between citizens and public institutions. Most legal and policy debates treat AI governance as a problem of automated decision-making: whether outputs are explainable, fair, reviewable or contestable. This article shifts attention upstream. It argues that civic platforms also exercise pre-decisional governance through interface architectures, ranking systems, moderation pathways and AI-supported clustering or summarisation. These mechanisms rarely decide a legal claim on their own, yet they determine which contributions become visible, comparable, aggregated and institutionally actionable. The article develops a conceptual framework for analysing this infrastructural layer of democratic authority. Methodologically, it uses an ideal-typical reconstruction of participatory platform workflows, informed by documented affordances of systems such as Decidim and Consul, rather than an empirical evaluation of a single deployment. The analysis connects platform governance and public-law accountability with the EU AI Act, the Digital Services Act, the GDPR and the Council of Europe AI Convention. It proposes four governance principles: visibility-mechanism transparency, procedural traceability, contestability by design and institutional responsibility mapping. The article concludes that democratic legitimacy in AI-mediated civic environments depends not only on accountable decisions but on accountable infrastructures of participation.