Detecting multi-modal GAI-manipulated tourism review

With the increasingly crucial role played by electronic word-of-mouth, online reviews have become an indispensable informational element for multiple stakeholders in the competitive tourism market. However, the rapid development of generative artificial intelligence (GAI) has not only threatened the unique position of humans as the sole producers of reviews but also broken new ground in the covertness and disorientation of manipulated reviews. To address this emerging issue, this study proposes a novel detection system, namely, multi-modal GAI-manipulated tourism review detector, which can accurately detect both textual and visual tourism manipulation through the innovative extraction of text linguistic features and image texture patch features. The superiority and effectiveness of the proposed detection system are demonstrated through empirical system application. This study not only offers theoretical and methodological references for tourism review detection research, but also contributes to the decision-making of tourists, the reputation of tourism enterprises and online travel agents.