Abstract

Like social media, AI has been widely associated with democratic decline, particularly in relation to the democratic information environment. Researchers point to the dangers posed by hyper-personalized AI content, AI-powered disinformation, political extremism and polarisation, declining trust in democratic institutions and a possible weakening of the state’s information capacity. At the same time, evidence is emerging that generative AI models can be helpful in improving democratic interaction (Summerfield et al, Costello et al, 2025, Tessler et al, 2024, Landemoore, 2023).

This talk examines some early evidence on the effects of AI on the democratic information environment. It reports research and introduces new projects on political persuasion; use of AI in the public sector; the nodality and legibility of government in the age of AI; information exchange between government and citizens; trust in AI as an information source; and competitive pressures for public attention. It also draws on the speaker’s work on ‘Digital Era Governance’ (Dunleavy and Margetts, 2006, 2023, 2024 and forthcoming in 2026). It proposes a research agenda for monitoring and measuring the democratic effects of AI going forward.