- Published on
AI summaries of Tripadvisor hotel reviews downplay serious complaints, investigation finds
- Authors
- Name
- DeepHorus
AI-generated summaries of Tripadvisor hotel reviews are downplaying serious complaints, including mass food poisonings and sexual harassment, according to an investigation by the consumer advocacy group Which?. The tool, designed to aggregate millions of reviews, has praised hotels with severe safety issues, such as one Cape Verde resort being sued for illness outbreaks, calling it 'spotless' and 'popular.' These summaries appear on hotel webpages to aid travelers in booking decisions, but critics argue they misrepresent customer experiences.
Which? examined multiple cases where the AI glossed over critical feedback. At the Riu Palace Santa Maria in Cape Verde, the summary highlighted spacious rooms and diverse restaurants, yet guests reported raw chicken, flies on buffets, and even 'dead little roasted mice.' The hotel chain is facing a high court lawsuit from hundreds of guests alleging poor hygiene. Similarly, a Dominican Republic hotel was praised for 'abundant' amenities despite guests showering with bottled water due to dry taps and widespread illness.
In a Turkish resort, the AI described service as 'friendly' despite numerous reports of sexual harassment from male staff. Guests stated they felt unsafe due to repeated inappropriate advances, including social media requests. Which? found that the AI summaries consistently sanitized negative feedback, focusing on positive aspects while minimizing serious warnings. Tripadvisor stated it is monitoring and refining the tool, but insisted it helps travelers quickly grasp feedback.
However, experts warn that AI summarization models tend to flatten criticism. Duncan Brumby, a professor at University College London, noted that AI often 'sanitizes and rubs off the edges' of complaints, likely due to training data dominated by bland observations. This can lead to misleading portrayals, as seen in these examples. Rory Boland of Which? Travel urged users to bypass summaries and consult actual reviews, especially one-star ratings, and cross-check with other sites.
Tripadvisor defended the feature, claiming it does not replace human reviews and that customers should use common sense. The platform also said it suppresses summaries when severe safety incidents like death or sexual assault are reported, but this mechanism failed in the cited cases. The findings echo other AI failures, such as Google's health summaries removed earlier this year after providing dangerous advice.
Moving forward, travelers are advised to scrutinize AI-generated overviews critically. While Tripadvisor refines its algorithms, the incident highlights broader challenges in relying on AI for nuanced consumer feedback. As Brumby suggested, these systems may inadvertently prioritize politeness over accuracy, underscoring the need for human oversight in travel planning.
