Perceptions of AI-generated news: an experiment on disclosure type and timing

Abstract

This study experimentally tested effects of AI disclosure (byline vs. late) on news perceptions (N=997). Preliminary results show upfront AI bylines (AI-only or AI-human collaboration) did not significantly impact credibility, readability, or usefulness compared to human-only. However, late disclosure of concealed AI use significantly reduced content and source credibility. Disclosure effects on engagement (clicking “read more”) were significantly moderated by news topic (politics, health, finance). Findings suggest transparency, more than AI use, shapes audience reactions.