TripAdvisor's AI Review Summaries Are Grossly Misleading
If you're using TripAdvisor's AI-generated summaries to decide where to book your next vacation, a new investigation suggests you might want to scroll past them and read the actual reviews.
An investigation by consumer campaign organization Which? found that TripAdvisor's AI overview tool — designed to summarize millions of customer reviews into quick, digestible highlights for hotel pages — consistently downplayed serious complaints, sometimes to the point of describing genuinely dangerous situations in glowing terms.
The most alarming example involves the Riu Palace Santa Maria in Cape Verde. The AI summary described it as popular, with spacious rooms, diverse restaurants that earn "rave reviews," and spotless cleanliness. What it didn't surface: customer reports of raw chicken being served, photographs of flies and birds on the buffet, and what one guest described as "dead little roasted mice by the sitting area." One reviewer wrote plainly: "This place will destroy holidays." The hotel chain is currently being sued in high court by hundreds of guests alleging illnesses linked to poor hygiene and food safety failures.
That AI summary has since been removed from TripAdvisor's site.
A separate Dominican Republic hotel was praised by the AI for "abundant" amenities, with only a vague mention of "inconsistent" cleanliness and "maintenance issues." Guest reviews told a different story — people were showering with bottled water because the mains water had stopped running, and half the guests at a large wedding party became sick.
Perhaps most troubling: a hotel in Turkey where guests reported repeated sexual harassment from male staff, including staff asking guests to connect on social media, was described by the AI as offering "friendly" service with only "lapses noted by a few."
Why AI Does This
The sanitizing effect isn't a glitch — it's a predictable outcome of how these systems work, according to researchers who study human-computer interaction.
"Here you have guests describing a really negative experience, but the AI has decided to tone it down," said Duncan Brumby, a professor at University College London. "It's as if it's being polite."
Brumby's own research into AI used in academic peer review found the same pattern — AI consistently "sanitises and rubs off the edges" of sharp criticism, likely because its training data contains far more neutral and positive language than strongly negative language. The math of the training set pulls everything toward the bland middle. A hotel with 800 mediocre reviews and 50 horrified ones gets summarized as if the 800 are the story.
Other research has found that AI tools designed to summarize consumer opinions systematically reduce the richness of feedback to shallower sentiments — a feature that works reasonably well when the goal is a quick vibe check, and catastrophically when the goal is safety information.
TripAdvisor's Response
TripAdvisor said it was "monitoring and refining" the tool and looking into the specific examples flagged. It maintained it was "confident these features are delivering exactly what they were designed to do: help travellers quickly understand the breadth of feedback while making it easy to explore the underlying reviews in full."
The company added that its systems automatically suppress AI summaries when reviews mention serious safety incidents including death, drugging, or sexual assault — though the harassment example flagged by Which? apparently didn't trigger that filter.
Rory Boland, editor of Which? Travel, was direct: "The platform has a responsibility to revisit the accuracy of its AI summaries. In the meantime, users should scroll past these summaries and look at guest reviews, particularly one-star ratings, and at reviews on other sites, to make sure their next stay is a safe one."
This Pattern Is Broader Than TripAdvisor
TripAdvisor isn't alone in discovering that AI summarization creates risks when applied to domains where the outlier negative experience is actually the most important information. Earlier this year Google removed some of its AI health summaries after a separate investigation found people were being misled by false and dangerous medical information generated by the same kind of summary tool.
The underlying problem is structural. AI summarization tools are optimized to surface what's representative — the typical experience, the average sentiment, the most common theme. But when you're deciding whether to take your family to a hotel or take a medication, the typical experience isn't what you need to know about. You need to know about the worst cases. And those, almost by definition, are the ones AI summary tools are least equipped to surface.
One-star reviews exist for a reason. Read them before you book.
Curious for more stories that keep you informed and entertained? From the latest headlines to everyday insights, YourLifeBuzz has more to explore. Dive into what’s next.