Production cues and conditional reliance in AI-mediated news: a three-study investigation of audience judgment under information asymmetry

As artificial intelligence (AI) becomes increasingly integrated into digital news presentation, understanding how audiences interpret AI-related production cues is critical to explaining evaluations of journalistic authority and reliance. Yet existing research has largely focused on transparency and generalized trust, leaving the relationships among production cues, perceived journalistic attributes, cognitive authority, and context-dependent reliance insufficiently understood. To address this gap, we propose a Signal–Attribute–Authority–Conditional Reliance (SAAC) framework and evaluate it across three complementary studies involving Chinese digital-news users: a 2 × 2 randomized experiment manipulating AI-disclosure and human editorial-review cues while holding the underlying AI-produced news stimulus constant ( n = 473); a cross-sectional survey integrating structural modeling, predictive analysis, and necessary condition analysis ( n = 522); and explanatory semi-structured interviews ( n = 29). The experiment showed that production cues influenced several post-exposure evaluations despite identical news content. Survey findings revealed associations linking perceived disclosure quality and human editorial oversight to perceived news quality and accountability, with further associations involving journalistic cognitive authority, conditional reliance, and acceptance intention, although explanatory and predictive performance remained modest. Interviews indicated that participants distinguished technological presentation capability from journalistic source qualification and evaluated reliance in relation to news topics, perceived stakes, editorial involvement, and institutional safeguards. Together, the findings suggest that,within the present study settings, AI-mediated news acceptance is better understood as an I nstitutionally situated and conditionally bounded judgment than as a simple function of transparency or generalized trust.