Trust at the Interface: Consumer Trust, Transparency and Persuasion Knowledge in AI-Mediated Marketing

Artificial intelligence (AI) now selects, personalises, writes and voices marketing communications, placing an algorithmic agent at the interface between firms and consumers. Yet the evidence on how consumers respond is fragmented across three largely separate literatures — studies of AI disclosure, of personalisation and privacy, and of algorithm appreciation and aversion — which imply conflicting managerial defaults. This paper integrates these literatures through a single interpretive mechanism drawn from the Persuasion Knowledge Model: persuasion-knowledge activation. We argue that interface design choices — whether AI involvement is disclosed, whether data collection is overt or covert, how humanlike the agent appears, and how the task is framed — do not act on trust directly. They act by changing what consumers infer about the persuader's goals and competence, an inference process in which privacy calculus, uniqueness neglect and identity threat also operate. The resulting Trust at the Interface framework links interface signals to consumer appraisals, to trust attributions directed separately at the AI system and at the firm, and to behaviour, and adds a failure-and-repair loop explaining asymmetric blame and recovery after AI errors. Seven formal propositions specify the framework's testable claims, with implications for interface design, consumer protection and future research.