Online consumer reviews function as trust infrastructure in e-commerce, yet the ecosystem is structurally vulnerable to manipulation. Generative artificial intelligence (GenAI) and large language models (LLMs) intensify this threat by lowering the cost of producing linguistically fluent, platform-native fake reviews, enabling paraphrase-based “review laundering,” and accelerating regeneration after takedowns. This review article provides a rapid, decision-oriented synthesis of GenAI-enabled fake reviews for marketplace researchers and practitioners. First, it establishes the pre-GenAI baseline threat model, highlighting why fake review markets persist under asymmetric information and imperfect monitoring. Second, it explains what is qualitatively different under GenAI, including cue inversion that weakens traditional text-based heuristics and the rise of paraphrased and hybrid human-AI review production. Third, it develops a GenAI-enabled attack typology and maps detection approaches by required data, strengths, and failure conditions, emphasising that text-only screening is increasingly insufficient under adversarial adaptation. Fourth, it proposes a layered “review integrity stack” that integrates policy clarity, risk-proportionate friction and provenance, multi-signal detection integrated with investigation workflows, credible enforcement, and transparency and redress. Finally, it outlines a focused research agenda that shifts the field from narrow classification performance to review integrity as a platform capability, with particular attention to emerging markets and cross-border commerce where institutional distance and multilingual variation can amplify harm. The paper offers actionable guidance for platforms and regulators seeking to preserve marketplace trust under machinegenerated persuasion.
Authors: Galvin Kuan Sian Lee
DOI: https://doi.org/10.6084/m9.figshare.32122843
Publish Year: 2026
