AI tools in Parliament: Analytical framework and empirical insights

Abstract Legislatures worldwide are increasingly developing artificial intelligence (AI) tools, yet political scientists still lack a shared framework for systematically comparing them. This article proposes a ‘Dual-Layer Model’ for analysing parliamentary AI tools, distinguishing between what these tools do and the actors involved in their use, that is, where inputs originate and where outputs are directed. Building on this framework, we develop a classification that captures how AI tools support different institutional functions within parliament, based on configurations of information flow between actors. This approach shifts attention from technical functions to how AI-generated outputs are embedded in relationships among parliamentary actors and between parliament and the public and provides a basis for studying the institutional consequences of AI adoption, the distribution of informational resources and capacities, and the strategic choices guiding the development of specific tools. Drawing on evidence from eleven parliamentary chambers, we show that the adoption of AI tools is currently shaped by an efficiency-led logic. Most tools are used to support the work of parliamentary staff and offices, consistent with earlier patterns of internet diffusion in parliaments, while tools used by legislators or incorporating citizen input remain comparatively limited. At the same time, patterns of adoption do not converge towards a common portfolio of AI tools across legislatures.