Quality Perceptions and Intended Engagement in Response to AI-Generated and AI-Assisted News

A survey experiment involving 599 participants in German-speaking Switzerland found that readers rate AI-generated and human-written news excerpts similarly in perceived quality, according to a preprint posted on arXiv [1]. The study also found that disclosing AI involvement temporarily increased readers’ willingness to continue reading an assigned article [1]. The preregistered experiment, led by Fabrizio Gilardi of the University of Zurich, asked participants to evaluate two short news excerpts before learning how each was produced [1]. The excerpts were either written entirely by a human, produced with AI assistance, or fully generated by AI [1]. Ratings of credibility, readability, and expertise did not differ significantly across the three conditions [1]. After the production method was disclosed, participants in the AI-assisted and fully AI-generated groups reported a higher willingness to continue reading the article they had been assigned, compared with a control group [1]. However, when asked about their future willingness to read AI-generated news in general, no difference emerged between conditions [1]. The findings suggest that while transparency about AI use can momentarily pique curiosity, it does not yet alter longer-term reading intentions [1]. The study arrives as newsrooms worldwide are integrating generative AI into editorial workflows. Automated journalism, also called algorithmic or robot journalism, now encompasses content production, data mining, news dissemination, and content optimization [4]. Since the 2020s, large language models built on the transformer architecture have enabled the generation and summarization of information at scale [5]. Advocates argue that automation can free journalists from routine reporting in sectors such as finance, sports, and election coverage, while critics raise concerns about transparency, accountability, and employment [4]. Gilardi, a Swiss political scientist and professor of policy analysis at the University of Zurich, has previously published on delegation theory and policy diffusion [9]. His latest work adds to a growing body of research examining how audiences perceive AI-generated journalism [1]. The paper was first submitted on 5 September 2024 and last revised on 17 June 2026 [1].