Can AI-Attributed News Challenge Partisan News Selection? Evidence from a Conjoint Experiment

Abstract

With artificial intelligence (AI) increasingly shaping newsroom practices, scholars debate how citizens perceive news attributed to algorithms versus human journalists. Yet, little is known about these preferences in today’s polarized media environment, where partisan news consumption has surged. The current study explores this issue by providing a comprehensive and systematic examination of how citizens evaluate AI-attributed news compared to human-based news from like-minded and cross-cutting partisan sources. Using a preregistered conjoint experiment in the United States (N = 2,011) that mimics a high-choice media environment, we find that citizens evaluate AI-attributed news as negatively as cross-cutting news sources, both in terms of attitudes (perceived trustworthiness) and behavior (willingness to read the news story), while strongly preferring like-minded sources. These patterns remain stable across polarizing and non-polarizing issues and persist regardless of citizens’ preexisting attitudes toward AI, political extremity, and media trust. Our findings thus challenge more optimistic views about AI’s potential to facilitate exposure to diverse viewpoints. Moreover, they suggest that increased automation of news production faces both public mistrust and substantial reader resistance, raising concerns about the future viability of AI in journalism.