The problem with trusting a machine to tell you about a hotel
Planning a vacation already feels like a second job. You scroll through pages of reviews, compare photos, check the star rating. Then along comes a neat little AI-generated summary that promises to do the heavy lifting for you. Sounds helpful, right?
Not so fast. An investigation by consumer group Which?, reported by The Guardian, found that TripAdvisor AI summaries are systematically filtering out the worst guest experiences — including hygiene failures, safety risks, and even sexual harassment.
In some cases, the gap between what the AI says and what real guests report is so wide it borders on dangerous misinformation.
Raw chicken, dead mice, and a cheerful summary
Take the Riu Palace Santa Maria in Cape Verde. The AI-generated overview called the resort “spotless” and said its restaurants earned “rave reviews.” Sounds like a dream.
Here’s what actual guests described: raw chicken served at the buffet, flies and birds buzzing around food stations, and dead mice near the seating area. The hotel chain is currently facing a High Court lawsuit brought by hundreds of guests who allege they fell ill due to hygiene failures at the property.
None of that — not the raw meat, not the rodents, not the lawsuit — appeared in the TripAdvisor AI summary. The company has since removed that particular overview.
When ‘friendly service’ means something else entirely
It gets worse. A hotel in Turkey had multiple guest reports of repeated sexual harassment by male staff members. The AI summary described the service as “friendly” with only “a few lapses.”
Think about that for a second. If someone booked that hotel based on the AI overview alone, they would have had no warning at all about the danger waiting for them. The summary didn’t just soften the criticism — it erased the severity of the complaint entirely.
These aren’t isolated glitches. The Which? investigation turned up multiple cases where serious safety incidents were either minimized or completely omitted from the automated summaries.
Why AI keeps sanding down the rough edges
Duncan Brumby, a professor of human-computer interaction at University College London, offered a straightforward explanation for why this keeps happening. Most of the data used to train AI language models leans toward polite, bland language. So when a guest writes an angry review about a dangerous situation, the model tends to treat it like a minor complaint.
Think of it this way: the AI has been trained on millions of sentences that say “the room was okay” and “the staff were nice.” It doesn’t have a strong framework for understanding that “I was sexually harassed” is not the same category of problem as “the Wi-Fi was slow.”
The result is a summary that flattens everything into the same polite, inoffensive tone — even when the original review is screaming about a genuine safety hazard.
What TripAdvisor says
TripAdvisor told the Guardian that it is investigating the mismatched summaries. The company says its systems are designed to suppress AI overviews when reviews mention serious safety incidents. It also stressed that these summaries were never meant to replace reading the actual reviews.
But that raises an obvious question: if the summaries are supposed to be suppressed when safety issues come up, why are so many dangerous complaints slipping through?
Don’t let an algorithm plan your next trip
None of this means you should stop using TripAdvisor entirely. The platform still contains a massive amount of useful, detailed feedback from real travelers. But the AI summaries? Treat them with serious skepticism.
Here’s a practical approach:
- Scroll past the AI summary. It’s usually at the top of the page. Ignore it and go straight to the written reviews.
- Sort by lowest rating first. The one-star and two-star reviews will tell you what actually goes wrong at a property. Read several of them to spot patterns.
- Check multiple platforms. Don’t rely on a single site. Cross-reference with Google Maps reviews, Booking.com, and Expedia to get a fuller picture.
- Look for specific details. A review that says “the bathroom was dirty” is less useful than one that says “there was mold in the shower and the toilet didn’t flush.” Specifics are harder to fake.
The same principle applies to AI-generated travel content more broadly. Whether it’s a hotel summary, a restaurant recommendation, or a suggested itinerary, algorithms are still terrible at understanding context, danger, and nuance. They can’t tell the difference between “the eggs were cold” and “I was served raw chicken.”
The bottom line on TripAdvisor AI summaries
AI tools are getting better at summarizing text. But better at summarizing is not the same as better at understanding. When it comes to something as personal and high-stakes as where you’ll sleep on vacation, the stakes are too high to outsource judgment to a machine that can’t tell the difference between a minor complaint and a major safety issue.
Read the real reviews. Ignore the AI gloss. And if a summary sounds too good to be true? It probably is.
