Abstract：In order to effectively identify and manage fake online reviews caused by artificial intelligence generated content (AIGC) in e-commerce platforms, we construct a hybrid BERT-BiLSTM-GAT model based on an active semi-supervised framework, propose an effective AI review recognition method, and empirically analyze the model in a variety of datasets and verify the model’s flexible recognition ability when dealing with multiple categories of product reviews, multiple AIGC models, and different prompt words. The results indicate that the method in this study achieves better recognition results in the down jacket dataset and cross-domain dataset, which is better than the current deep learning recognition methods, and verifies the performance advantage of the model. At the same time, in the case of high consistency between the features of the AI review and the manual review, it still possesses good generalization ability and adaptability.
李世勇，杨铮铮. 基于BERT-BiLSTM-GAT的人工智能生成电商虚假评论识别研究[J]. 管理学报, 2025, 22(3): 557-.
LI Shiyong，YANG Zhengzheng. Identification of E-Commerce Fake Reviews Generated by Artificial Intelligence Based on BERT-BiLSTM-GAT. Chinese Journal of Management, 2025, 22(3): 557-.
