The Moderating Role of Anthropomorphic Presentation in Audience Recognition of Generative AI News: A MAIN-Based Model and Experimental Protocol

Abstract

Humanlike news interfaces can make generative artificial intelligence appear socially familiar while leaving the origin of a news report uncertain. This article develops a conditional process model of audience recognition of AI authorship, drawing on the Modality, Agency, Interactivity, and Navigability model. It distinguishes recognition of production origin from perceived credibility, social presence, and identification with a presenter. Truthful authorship disclosure is specified as the predictor, anthropomorphic presentation as a randomized moderator, and source scrutiny and social presence as parallel mediators. A prospective factorial experiment crosses disclosure with presentation style and includes both AI-drafted and human-written news. The proposed analysis separates AI-attribution judgments from classification accuracy and examines false-positive judgments for human-written reports. It uses crossed participant–stimulus models for randomized treatment effects and participant means for explanatory conditional indirect associations. Competing predictions are retained where existing evidence does not justify a directional hypothesis: a disclosure may encourage source scrutiny or make further scrutiny unnecessary, while humanlike presentation may either compete with or heighten attention to provenance cues. The article provides operational definitions, a stimulus protocol, draft measures, and an analysis plan; it reports no new participant data or completed mediation test. Its contribution is a testable account of how interface design may condition the relationship between disclosure and recognition without treating social engagement as evidence of mistaken authorship attribution.