Abstract

Online platforms increasingly use AI to generate summaries of consumer reviews, offering efficiency but raising concerns about distortion and bias. This research-in-progress paper examines deception in AI-generated review summaries, conceptualized as systematic mismatches between what reviews express and what summaries communicate. The paper proposes a structured framework that identifies five types of deception—omission, exaggeration, framing bias, hallucination, and lack of responsiveness. Each type is grounded in communication theory, natural language processing research, and information systems scholarship, and is illustrated through measurable indicators such as coverage of review aspects, sentiment divergence, linguistic framing, evidence support, and timeliness of updates. Building on this framework, the paper discusses directions for expansion and empirical verification, including when these deceptive patterns arise and how they influence user perceptions and platform trust. Overall, the study provides a preliminary foundation for examining AI-driven misrepresentation and considers possible implications for transparency and consumer protection in digital marketplaces.

Download

Share

COinS