Synthesizing agendas: mapping the transformative potential of generative AI for agenda setting effects

Abstract From an agenda-setting perspective, the present study compares the structure of issue, candidate, and party emphasis in AI chatbot responses during the 2025 Norwegian election campaign with content from search engines and news websites. The study shows that when users employ conventional queries in AI chatbots, the issue and candidate saliences are very similar in chatbots, search engines, and editorial news media content; the main difference is that AI chatbots give roughly similar attention to all major parties rather than focusing the large parties like the news media. When using less conventional queries that feed personal information, e.g. about one’s occupation or ideology, the AI chatbots significantly shift their focus and emphasize a query-specific set of issues and parties. An accompanying small-scale voter survey finds that current impact of AI chatbots is low due to limited user adoption and dominance of conventional prompts that would lead to issue, party, and candidate saliences similar to those in the editorial news media. The already visible potential for users and chatbots to co-create an entry point into self-reinforcing spirals of partisan selectivity warrants attention. Further normative implications of this potential are discussed.