The Power of Conversation: A Survey Experiment on AI-Generated Information and Political Behaviour

Citizens increasingly seek political information from generative AI platforms such as ChatGPT. Despite concerns about the reliability of these platforms, extensive usage by citizens of these tools is poorly understood, including implications for political knowledge, voter behaviour, and political trust. Crucially, these platforms differ from news, social media, and search engines, in their interactivity and human-like tone. These unique characteristics may increase citizens' trust in the information provided and affect perceptions of political knowledge and efficacy. Yet, there is little causal research on the relationship between AI platforms and political behaviour. We examine the effects of AI-produced information on citizens’ political behaviour in the context of the European Parliament – characterised by low baseline knowledge and trust, high complexity, and weak partisanship. We field a preregistered survey experiment, with live participant interaction with a novel generative-AI transparency platform, varying two attributes: (1) interactivity (ability to ask follow-up questions) and (2) response tone (friendly, human-like vs. neutral), holding substantive content constant. We test whether AI’s unique conversationality increases citizen’s trust in the information provided, AI generally, as well their political knowledge, efficacy, and institutional trust.