Key Facts

• TripAdvisor uses AI to generate summaries of hotel reviews, but an investigation by consumer group Which? revealed these summaries omit serious issues.
• Examples include the Riu Palace Santa Maria in Cape Verde where AI called it "spotless" while guests reported raw chicken, flies, dead mice, and a lawsuit over hygiene failures.
• A hotel in Turkey saw guests report repeated sexual harassment from staff, but the AI summary called the service "friendly with only a few lapses."
• Professor Duncan Brumby (UCL) explains AI models sand down negativity because training data is often bland and polite, making serious complaints sound like minor inconveniences.
• TripAdvisor says it suppresses summaries when serious safety incidents are mentioned, but the investigation shows gaps.

Planning a trip is stressful enough without wondering if the glowing hotel summary you just read was written by an AI that skipped the scary parts. As it turns out, that might be exactly what's happening on TripAdvisor.

According to an investigation by consumer group Which?, reported by the Guardian, TripAdvisor's AI-generated review summaries are smoothing over serious guest complaints — and in some cases, downright dangerous ones. The implications for travelers are significant: a glowing AI summary can mask hygiene failures, safety risks, and even criminal behavior, potentially leading to ruined vacations or worse.

Can You Trust an AI Summary of Your Hotel Reviews?

Take, for example, the Riu Palace Santa Maria in Cape Verde. The AI summary called it "spotless," with diverse restaurants earning rave reviews. Real guests, on the other hand, described being served raw chicken, flies and birds near the buffet, and even dead mice by the seating area. The hotel chain is currently being sued in the High Court by hundreds of guests over alleged illnesses linked to hygiene failures. You wouldn't know any of that from the AI's cheerful summary, which has since been taken down.

If you thought that was bad, consider a hotel in Turkey where guests reported repeated sexual harassment from male staff. The AI summary called the service "friendly with only a few lapses." If someone booked the hotel based on the AI overview of user reviews, they would have no inkling of the horror they could face there. Several other examples found by the investigation paint a consistent picture: AI summaries are flattening critical feedback into harmless platitudes.

The problem is not limited to TripAdvisor. Other platforms like Google, Amazon, and Yelp have also deployed AI-generated summaries for user reviews, with similar issues emerging. For instance, Google's AI overviews for restaurants have occasionally missed negative health inspection reports. The core challenge is that natural language processing models tend to average out sentiment, giving undue weight to the majority of neutral or positive reviews while downplaying outliers — even when those outliers contain life-saving information.

Why Does AI Keep Softening Bad Reviews?

A UCL professor of human-computer interaction, Duncan Brumby, offered a simple explanation. AI models tend to sand down harsh criticism because most of their training data leans bland and polite. When a guest writes a negative review, the AI sometimes treats it like a minor inconvenience — a complaint about noise becomes "some guests mentioned street sounds," while a report of sewage backup becomes "occasional plumbing issues." The model's objective to produce coherent, neutral-sounding text often conflicts with the factual accuracy of extreme cases.

This smoothing effect is compounded by the way review data is aggregated. AI summaries typically extract the most common phrases and themes. If 90% of reviews for a hotel are positive but 10% mention serious safety violations, the model may interpret the violations as statistical noise and omit them. The underlying architecture of large language models — which rely on probabilities from training data — lacks an inherent understanding of severity. A sentence about a broken air conditioner and a sentence about a bedbug infestation may be treated as equally negative, even though one is a minor annoyance and the other is a public health hazard.

Historical Context: The Long History of Review Manipulation

AI summarization is only the latest chapter in a long saga of unreliable online reviews. Long before generative AI, TripAdvisor and other platforms fought fake reviews, paid placements, and biased filtering. In 2018, a study found that up to 20% of TripAdvisor reviews were potentially fraudulent, leading the company to implement stricter verification measures. However, AI summaries introduce a new layer of opacity: even when individual reviews are genuine, the algorithm's interpretation can distort the truth. This is especially dangerous because users tend to trust AI-generated content as objective, a phenomenon known as automation bias.

TripAdvisor says it's looking into the mismatched summaries and that its systems suppress AI overviews when reviews mention serious safety incidents. It also maintains that these summaries were never meant to replace actual reviews. However, the company's response underscores a broader industry problem: AI tools are deployed to improve user experience, but guardrails fail when content is ambiguous or when malicious actors exploit the system. Even if TripAdvisor improves its model, the underlying tension remains — a summary by definition sacrifices detail for brevity, and when the omitted details involve safety, the trade-off is unacceptable.

What Travelers Can Do to Protect Themselves

The takeaway is simple. Don't let AI summaries make your travel decisions for you. Scroll past them, read the one-star reviews, and check other sites too. Look for patterns: if multiple recent reviews mention cleanliness issues, staff behavior, or food quality, take them seriously. Cross-reference with dedicated travel forums, social media, and local news reports. AI summaries have time and again proven to be fraught with inconsistencies, and it's always good to do your own research — lest your vacation turns into a nightmare.

Beyond individual action, there is a need for regulatory scrutiny. Consumer protection agencies in the UK and EU are already examining how AI-generated content interacts with existing consumer laws. In the US, the Federal Trade Commission has warned about deceptive practices in online reviews. As AI summarization spreads, authorities may need to set standards for transparency — for example, requiring platforms to clearly label AI-generated summaries and provide a confidence score, or allowing users to opt out of seeing them entirely. Until safeguards are in place, critical thinking and manual verification remain the traveler's best defense.

The investigation by Which? has already led to changes: some of the most misleading summaries have been removed from TripAdvisor. But the underlying issue is systemic. As AI becomes the default interface for processing human feedback, every industry that relies on user reviews — from hospitality and dining to healthcare and education — faces the same challenge: how to condense many voices into a few sentences without losing the signal of warning. For travelers, the lesson is clear: an AI summary is a starting point, not a final verdict. Always dig deeper.

Source: Digital Trends News