Tripadvisor's AI Summaries Called a Hotel "Spotless." Guests Were Suing Over Dead Mice.

So here's what actually happened. A consumer group in the UK pulled up AI-generated summaries on Tripadvisor for hotels with documented, serious problems... we're talking active lawsuits, guests reporting illness, staff harassment... and found the AI cheerfully describing these properties as "spotless" and "friendly." One hotel had hundreds of guests suing the chain in high court over food safety. The AI summary? Clean. Wonderful. Come on in.

Let me run that through the Dale Test for a second. What is this system actually doing? It's ingesting thousands of reviews, many of which describe genuinely dangerous conditions, and producing a summary that reads like marketing copy. This isn't a bug. This is what large language models do when you point them at mixed-sentiment text without explicit instructions to weight severity. They average. They smooth. A hundred "nice pool" reviews and three "I found a rodent in my food" reviews become "guests enjoyed the amenities with occasional notes about dining." That's not a summary. That's a cover-up generated at scale. And the worst part? Tripadvisor says they "fundamentally disagree with the premise" of the investigation. Which tells you everything about how seriously they're taking it.

Here's what's actually broken. Tripadvisor's own data says 10.7% of reviews on their platform in 2024 were likely AI-generated... up 137% from 2019. So you've got an AI summarizing reviews that are increasingly written by AI. If that doesn't make you pause, I don't know what will. The signal-to-noise ratio was already getting worse before they layered a summarization model on top. Now the noise has a friendly face and a confident tone. Meanwhile, the operators who are actually investing in quality... training staff, fixing problems, responding to complaints... are competing against properties whose real issues are being algorithmically buried. That's not a level playing field. That's the platform picking winners by accident.

Look, I get why Tripadvisor did this. They're losing organic search traffic (some estimates put the decline at 33% from AI search tools answering queries directly). They need to keep users on-platform, and quick AI summaries are a retention play. But the execution is irresponsible. When your summary engine describes a hotel where guests report sexual harassment from staff as having "friendly" service, you don't have a product refinement issue. You have a liability issue. And you have an integrity issue that undermines the entire premise of user-generated reviews. Which? Travel's advice to users was blunt: scroll past the AI summaries entirely and read the one-star reviews yourself. When a consumer advocacy group is telling people to ignore your core feature, your product has failed the most basic test... does it help the user make a better decision?

What this actually does for independents and quality-focused operators is quietly devastating. Reviews were supposed to be the equalizer. The 90-key independent that delivers a genuinely better guest experience should outperform the 300-key branded property cutting corners. That was the promise. But if the AI summary flattens the difference... making a dangerous property sound "mostly fine" and a great property sound "mostly fine"... then the competitive advantage of actually being good gets erased by an algorithm that doesn't know the difference between "the towels were thin" and "I was harassed by a staff member." Those are not equivalent data points. But to a language model averaging sentiment across thousands of reviews, they're both just negative tokens getting diluted by volume.