An investigation has revealed that artificial intelligence-generated summaries on a major travel review website have minimized serious complaints reported by hotel guests, including instances involving food poisoning, poor hygiene, and sexual harassment. The AI-generated overviews, designed to distill millions of individual reviews into concise summaries, have been criticized for underrepresenting critical safety and service issues.
One notable example involved the Riu Palace Santa Maria in Cape Verde, where hundreds of guests have filed a legal suit alleging foodborne illnesses linked to inadequate hygiene and food safety practices. Despite these allegations and reports of raw chicken, flies, birds on the buffet, and even dead rodents near seating areas, the AI summary described the hotel as “spotless,” praised its “diverse restaurants,” and cited “rave reviews.” The hotel chain, RIU Hotels & Resorts, emphasized its commitment to high standards of professionalism and hygiene following the publication of the AI summary.
Similarly, a resort in the Dominican Republic had an AI-generated review highlighting its “abundant” amenities and only briefly mentioning “inconsistent” cleanliness and “maintenance issues.” Meanwhile, guests documented situations such as the unavailability of mains water, resorting to bottled water for showers, and widespread illness among visitors. Another problematic case involved a hotel in Turkey where guests reported feeling unsafe due to sexual harassment by male staff, including inappropriate social media requests. The AI summary, however, described the service as “friendly” and referred only to “lapses [in service] noted by a few.”
In response, the travel site’s operator stated that it is actively monitoring and refining the AI tool, affirming its confidence that the summaries serve their intended purpose—providing travelers a quick overview of guest feedback while encouraging review exploration in full. The company also said that the AI-generated content does not replace original traveler reviews and that summaries are automatically suppressed for properties associated with serious safety incidents such as death, drugging, or sexual assault.
Consumer group Which? urged caution, with Rory Boland, editor of its travel section, recommending that users bypass AI summaries and focus more on detailed guest reviews, especially low-rated ones, as well as reviews on other platforms to ensure informed decisions. Experts in human-computer interaction also highlighted risks inherent in AI interpretation. Duncan Brumby, a professor at University College London, noted that AI models tend to “sanitize” or soften sharp criticisms, possibly due to their training on predominantly neutral or mild language, resulting in summaries that downplay guests’ negative experiences.
As AI tools become increasingly integrated into consumer-facing platforms, the balance between brevity and accuracy remains a challenge, with this case underscoring the potential consequences of relying on automated summaries for critical service evaluations.