AI-Generated News and Its Impact on Public Trust in Journalism

Abstract

Generative AI is reshaping how news is produced and encountered, raising fundamental questions about public trust. This study develops a hybrid credibility framework that distinguishes a process-legitimacy path (disclosures about how stories are made) from a provenance-credence path (verifiable signals of media origins and edits). We report results from a preregistered 2×2×2 between-subjects experiment (Authorship: AI vs. human; Label: present vs. absent; Platform: publisher site vs. social feed; N=640) with story type blocked. AI authorship produced a modest trust penalty, larger in feeds and for hard news. Labels substantially increased perceived transparency and slightly improved perceived accuracy, but also carried a small direct negative effect on trust, consistent with algorithm-aversion heuristics; the net effect was positive among high-AI-literacy participants and negative among low-literacy participants. Parallel-mediation analyses showed that perceived accuracy and perceived bias transmitted authorship effects to trust, while transparency mediated label effects. We provide design guidance for contextual disclosure wording (“AI-assisted and editor-reviewed”), layered provenance that travels with assets, and visible editorial oversight across devices and distribution channels online. We conclude that calibrated trust in AI-augmented journalism depends on pairing usable transparency with verifiable provenance and audience literacy supports.

Keywords: AI-Generated News; Trust In Journalism; Source Credibility; Transparency; Provenance; Ai Literacy; Social Platforms.