Date of Publication : June 2026

Abstract: Online reviews play a crucial role as they influence the buying choices of customers in digital platforms. However, there has been a steep rise in fake and misleading reviews which reduces the reliability of such systems, leading to work on developing automated methods to detect them. Over the last ten years, this field of fake review detection has progressively evolved from simple linguistic and psycholinguistic features to behavioral analysis, network-based models, deep learning and the recent transformer-based techniques. Despite these advancements have perceived high accuracy on benchmark datasets, there are still limitations that holds back from real-world deployment. In particular, the existing models shows difficulty in adapting to changes over time (limited robustness), poor cross-domain generalization and in integrating semantic (text-based) and relational (network-based) detection paradigms. Moreover, with large language models in usage, the ai-generated fake reviews have been introduced which propose challenges by making linguistic deception cues unreliable. This paper provides a comprehensive review of fake detection methods, along with their evolution and critically analyzing unresolved challenges that exists across different approaches. By analyzing insights from prior foundational researches and studies, deep learning models, network-based methods, and recent surveys, we focus on outlining existing research gaps and suggest future directions to develop more robust, reliable and adaptable fake review detection systems.

Reference :