Fake Review Detection: Taxonomies, Benchmarks, and Intent Modeling Frameworks

Fake reviews on digital platforms are proliferating to a great extent, which is a challenge to the integrity of digital consumer ecosystems. This has put the development of reliable detection mechanisms as a critical research priority in influencing purchasing behavior and business reputation. In this paper, the rule-based feature-engineered machine learning systems, deep neural networks, transformer-based architectures, and emerging intent-aware frameworks are comprehended concerning fake review detection. These approaches are then classified with respect to model structure, feature dependence, and domain adaptability introduced with the help of a structured taxonomy. This study compares existing methods and uses comparative analysis to identify key limitations including domain sensitivity, overreliance on textual content, lack of interpretability, and low adversarial robustness. It particularly focuses on detection strategies in the light of modeler intent while respecting reviewer behavior using persona-based architectures and contrastive embedding alignment. These approaches allow for zero-shot detection as a path to more generalizable and semantically grounded systems. The paper also brings out the need for standardized benchmarking practices, ethically sourced datasets, and interdisciplinary methods incorporating natural language processing, behavioral analytics, and explainable AI. This work aids in bringing these scalability, transparency, and ethical alignment solutions a step closer, by critically evaluating the current landscape and proposing new research trajectories for fake review detection.