What the paper studied
Hotels increasingly use AI-enabled services such as chatbots, automated check-in/check-out, recommendation systems and intelligent self-service applications. This study asks which perceptions of those services shape tourists' behaviour afterwards, specifically whether they share positive or negative online reviews. The authors extend the Model of PC Utilization (MPCU) with constructs from the Motivational Model and UTAUT2, plus hospitality-specific factors: perceived ease of use, operational improvement, need for human interaction, perceived risk, perceived intelligence and trust. Data came from a structured questionnaire completed by a convenience sample of 400 tourists staying at a four-star hotel on a Greek island in summer 2025, analysed with binary logistic regression.
Key findings
- Perceptions of AI services predict review behaviour. Tourists' evaluations of AI-enabled services show differentiated associations with positive and with negative online review sharing.
- Five factors stand out. Ease of use, perceived intelligence, trust, perceived risk and the need for human interaction play key roles in shaping that behaviour.
- AI touchpoints feed electronic word of mouth. The study links AI service evaluations directly to post-stay electronic word of mouth (eWOM), an outcome earlier adoption research has rarely measured.
Why it matters for hospitality
AI-enabled services are usually justified by efficiency, but this study shows they also shape what guests say about a hotel online. A clumsy chatbot or an untrusted automated check-in can become a negative review, while a smooth, intelligent experience can become a positive one, with direct effects on reputation and future bookings.
Practical takeaways
- Prioritize ease of use and visible intelligence in guest-facing AI; these perceptions are tied to what guests later share online.
- Build trust and address perceived risk in AI touchpoints, since both shape review behaviour.
- Keep a human option available for guests who value human interaction. The authors recommend combining trustworthy, user-friendly AI with personalized human service.
Limitations
The sample is a convenience sample from a single four-star hotel on one Greek island during one summer season, so results may not generalize to other markets, hotel types or seasons. The design is cross-sectional and self-reported, and logistic regression identifies associations rather than causal effects. This summary is based on the article's published abstract.