A hotel currently facing a lawsuit due to multiple food poisoning incidents was labeled as “spotless,” while a resort where guests reported instances of sexual harassment by staff received commendations for its “friendly” service, all based on an AI tool designed to analyze vast amounts of Tripadvisor reviews.
An investigation by consumer advocacy group Which? revealed that these AI-generated summaries often glossed over serious issues, such as pervasive mold odors and inadequate water supply. These summaries appear on hotel listings to assist travelers in making informed booking decisions. For instance, the AI's description of the Riu Palace Santa Maria in Cape Verde highlighted its popularity, spacious accommodations, “diverse restaurants” with “rave reviews,” and immaculate cleanliness.
Contrarily, guests shared alarming reports of being served undercooked chicken, encountering insects around the buffet, and even discovering small deceased animals in common areas. One family, whose members fell ill during their stay, lamented, “This place will destroy holidays.”
The hotel chain is currently embroiled in a high court lawsuit brought by numerous guests who are alleging that health issues stemmed from poor hygiene and insufficient food safety practices.
Although the AI summary for the Cape Verde hotel has since been removed, RIU Hotels & Resorts stated that they operate with the utmost professionalism and prioritize hygiene and safety standards.
Another hotel in the Dominican Republic was commended by the AI for its plentiful amenities, with only a brief acknowledgment of “inconsistent” cleanliness and various maintenance problems. Guests reported needing to use bottled water for showers due to the lack of running water and that many members of a wedding group fell ill.
In Turkey, patrons commented on feeling unsafe because of ongoing sexual harassment from male staff members, including unsolicited requests to connect via social media. Nevertheless, the AI description characterized the service as “friendly,” only mentioning minor “lapses” in service, as uncovered by Which?
Tripadvisor acknowledged that it is continuously monitoring and improving its AI tool, stating its belief that these features successfully fulfill their intended purpose: helping travelers quickly grasp a wide spectrum of feedback while easily accessing complete reviews.
The company emphasized that AI-generated summaries do not replace user reviews, encouraging travelers to cross-check any AI recommendations against the extensive pool of over a billion reviews available on their platform.
Rory Boland, editor of Which? Travel, remarked on the platform's obligation to reassess the accuracy of its AI summaries and chatbot. He urged users to bypass these AI-generated overviews in favor of delving into actual guest feedback, particularly one-star reviews, and to explore other review platforms to ensure safe travel arrangements.
Tripadvisor added that its systems automatically hide AI summaries in cases of severe safety warnings, such as incidents involving death or sexual assault, to ensure this critical information remains highly visible to its audience.
Duncan Brumby, a human-computer interaction professor at University College London, noted that his research aligns with these findings, highlighting how AI often tends to “sanitize” explicit critiques, likely due to the prevalence of more moderate training data.
He observed, “Guests share distressing experiences, yet the AI neutralizes these accounts, as if it's being overly courteous.” Studies have also suggested that opinion-summarizing technologies frequently overlook the nuances of consumer feedback, diminishing it to simplistic sentiments.
Earlier this year, Google retracted some of its AI health summaries after a Guardian report revealed potential harm to individuals due to the dissemination of misleading information.
