Tripadvisor’s AI review summaries are drawing sharp criticism after a consumer watchdog found the feature glosses over serious hotel complaints, including food poisoning outbreaks, hygiene failures and sexual harassment allegations.
Consumer group Which? published the findings this week, and the results raise an uncomfortable question for travelers. Can a quick summary help people book faster without burying the warnings they actually need?
Tripadvisor built its AI review summaries to condense thousands of guest comments into a short paragraph that appears near the top of each hotel page. The idea is simple: scan the highlights and skip the scrolling. But Which? said several of these summaries told a far rosier story than the reviews sitting directly underneath them. They were blatantly inaccurate.
A five-star hotel called “spotless”
The clearest example centers on Riu Palace Santa Maria, an all-inclusive resort in Cape Verde. Tripadvisor’s AI review summary called the property “popular with many travellers,” praised its “spacious rooms” and “diverse restaurants” that earn “rave reviews,” and described cleanliness as “spotless.”
Recent guest reviews tell a different story. Which? said it counted 102 mentions of food poisoning at the resort as of March, along with 32 one- and two-star reviews posted between December 2025 and April 2026. Fourteen of those reviews said at least one person in the traveling party fell seriously ill. One guest wrote a blunt warning: “This place will destroy holidays.”
The hotel now faces a group legal claim on behalf of at least 412 holidaymakers who say they became sick after staying there, and Which? cited seven reported deaths linked to the property since 2023. RIU Hotels & Resorts disputed any suggestion of unsafe conditions, saying guest health remains a top priority and that its Cape Verde properties meet international hygiene standards.
Guests also raised an alarm about staff conduct. Two separate reviewers said a male staff member approached their daughters to ask for social media contacts, and one described being followed up a staircase toward her room. Tripadvisor’s AI review summary described the service simply as “friendly,” adding only that “lapses noted by a few” had occurred.
Chatbot called illness risk “unlikely”
Which? also tested Ollie, Tripadvisor’s AI trip-planning chatbot, by asking it directly about food poisoning risk at the same resort. Ollie called the chance “quite unlikely” and said the hotel had “a strong reputation for high hygiene standards,” according to the investigation.
That answer matters because more travelers now lean on quick AI tools instead of reading dozens of reviews themselves. When a chatbot waves away a documented illness pattern, it can steer a family straight into danger rather than away from it.
Which? found comparable gaps elsewhere. At Garza Blanca resort in Cancun, Mexico, multiple guests, including members of a wedding party, reported getting sick within the past year, yet the AI review summary highlighted “immaculate cleanliness.” At Occidental Caribe in the Dominican Republic, one recent visitor called the stay a “disaster and disturbing,” while another labeled it the “worst place imaginable,” citing sewage smells, mould and water shortages. The summary there pointed instead to “abundant” amenities.
At Kaia Coracesium on Turkey’s Antalya coast, several guests said male staff made them feel unsafe through repeated harassment and unwanted requests for contact information. Once again, the AI review summary settled on the word “friendly.”
Tripadvisor pushes back
Tripadvisor rejected the investigation’s central claim.
“We fundamentally disagree with the premise of this investigation,” a company spokesperson said, adding that AI review summaries draw from large volumes of user-generated content and never replace individual reviews.
The company said travelers can click through to the specific quotes behind each summary point.
Tripadvisor also said built-in safeguards suppress AI review summaries entirely for listings that carry traveler warnings involving death, drugging, or sexual assault, and it maintains that no review content vanished because of the tool. Ollie, the company added, remains a product still in development.
A trust problem on two fronts
Tripadvisor has already battled skepticism over fake reviews; its own 2025 transparency report said the platform removed more than 214,000 reviews in 2024 suspected of containing AI-generated text. Now the concern runs the other direction, toward whether its AI review summaries fairly represent the reviews that remain.
For now, consumer advocates recommend a low-tech workaround: read the lowest-rated reviews first, search for repeated red flags like illness or harassment, and cross-check recent posts on more than one site before booking.
Do you trust AI review summaries when booking a hotel, or do you still read the reviews yourself? Please share your views below.