Seeing AI in the Byline: Disclosure Labels, Visual Scrutiny, and Trust in AI‐Attributed News

ABSTRACT

AI disclosure labels may influence not only whether readers trust news but also how they inspect it. Drawing on the Heuristic‐Systematic Model and source‐credibility research, this study distinguishes overall news credibility, production–process trust (confidence in the accuracy and consistency of the information‐production process), and media trust (confidence in editorial judgment and accountability). A 2 × 2 laboratory experiment with a university‐community sample ( N  = 412) compared AI‐attributed and human‐attributed closely matched article versions across high‐ and low‐salience topics while recording gaze behavior with a Tobii Pro eye‐tracker. AI‐attributed versions received lower overall credibility and media‐trust ratings, higher production–process trust, and longer fixations and more re‐fixations on factual‐claim regions. Readers also spent more time on the AI disclosure zone and returned to it more often, while supplementary‐link clicking did not decline. Topic salience amplified the credibility and media‐trust deficits. The gaze pattern shows that attribution redirected attention toward factual claims and is consistent with selective scrutiny, although it does not demonstrate successful verification. Because the source conditions used closely matched rather than identical article versions, the inferences concern AI‐attributed versus human‐attributed stimuli under controlled equivalence checks, not an isolated label‐only effect. The findings support disclosure that specifies what the system did, what humans verified, and who retained editorial responsibility.