Synthetic voices and mediated trust: a multimodal analysis of AI-generated political communication in Indonesia

Abstract

Synthetic political communication increasingly detaches recognisable voices and embodied identities from biological speakers, complicating how authority, authenticity, and evidentiary credibility are established within platformed electoral environments. This study examines how AI-generated political communication in Indonesia constructs mediated trust through synthetic voice, multimodal orchestration, disclosure practices, and institutional contestation. A qualitative multimodal corpus design analysed seventeen publicly verifiable documents collected between January 2023 and June 2026, comprising eight primary artefacts, four secondary verification sources, and five institutional context documents through three integrated coding instruments. Findings show that seven primary artefacts combined synthetic speech with recognisable political likenesses, while public or direct address constituted the dominant communicative configuration. Political authority represented the most recurrent trust cue, whereas linguistic revoicing, historical revival, material promises, and apparent private access provided alternative routes to plausibility. Seven artefacts lacked clear AI disclosure, and every primary case subsequently attracted verification, warning, denial, or documentation from governmental, electoral, journalistic, or fact-checking institutions. This study introduces an integrated framework that connects vocal identity, cross-modal congruence, platform affordances, trust cues, disclosure, communicative function, and institutional response, thereby conceptualising mediated trust as a transferable yet contested political resource rather than a measurable audience attitude within contemporary digitally mediated democratic communication systems.