Which democracy does AI support? Mapping AI applications for civic participation

Artificial intelligence (AI) is increasingly implemented to support civic engagement and participation. However, there remains a lack of systematic analysis regarding what AI applications for civic participation do and, more generally, the type of democracy they rely on. The article addresses this gap by presenting a mapping of 77 AI-based civic participation tools identified through expert interviews and desk research across repositories of civic tech initiatives. Using Dahlberg’s typology of four positions of digital democracy (aggregative, deliberative, contestationary, and commons), the study evaluates the democratic affordances embedded in these applications. The mapping reveals a dominance of unitary models of democracy: 47 tools align with aggregative democracy and 26 with deliberative democracy, while agonistic democracy is hugely underrepresented, with just 4 contestationary applications and 0 projects oriented towards commons democracy. AI embodies an “anti-agonistic” script that prioritizes consensus and information processing over political conflict. I argue that this trend is driven by three key factors: the current hegemony of deliberative paradigms in democratic innovation, an AI “script” which favor sensemaking and information processing, and the marketization of civic tech. The article concludes that by sidelining the agonistic dimension of politics, current AI applications risk reducing democratic engagement to a matter of efficient data management, with limited opportunities for political action and power redistribution.