Which? Investigation: Tripadvisor's AI Summaries Mask Food Poisoning and Harassment

British consumer group Which? published an investigation on July 3 revealing that AI-generated review summaries on Tripadvisor can portray dangerous hotels as spotlessly clean and safe. Which? editor Rory Boland summed up the phenomenon in blunt terms:

Potentially life-threatening. - Rory Boland, editor, Which?

What the investigation found

Analysts compared the content of AI-generated review summaries with actual guest reviews across more than a dozen properties. At the Riu Palace Santa Maria in Cape Verde, which faces a class-action lawsuit from more than 400 affected tourists and has recorded seven deaths since 2023, Tripadvisor's algorithm summarized reviews with the word "immaculate" and referred to "enthusiastic reviews." Yet the guest reviews themselves contained 102 mentions of food poisoning, and 14 serious illness cases were documented between December 2025 and April 2026.

A similar pattern repeated at other properties. At Mexico's Garza Blanca, where an entire wedding party fell ill, the AI assured readers of "immaculate cleanliness." At the Dominican Republic's Occidental Caribe, guests reported no running water and sewage smells, while the summary spoke of "rich amenities" and "minor shortcomings." The most troubling example involves Turkey's Kaia Coracesium, where numerous women reported sexual harassment by staff, which the algorithm framed as "minor service issues."

Rivals do it better

Which? points out that the problem isn't AI technology itself, but Tripadvisor's specific implementation. For the Britannia hotel chain in London, Google's summaries accurately reflected guest complaints about "dirty" rooms and "terrible" service, while Tripadvisor, over the same period, described rooms guests "often praise for cleanliness" and a "charming" atmosphere. This shows that other AI systems can pick up on the same warning signs that Tripadvisor's algorithm misses.

Which?'s key criticism concerns not just the content but where it's displayed. The AI summary appears at the very top of a property's page, before users even reach individual reviews. Rory Boland stressed that this placement is precisely what turns the omission of critical safety information into a real danger, since some travelers never scroll further down.

Tripadvisor's response

Tripadvisor disputed the report's conclusions, saying its AI summaries are designed "to maintain the honesty and transparency" on which the company has built customer trust for more than 25 years. The company also noted that its AI tool, named Ollie, is still "a product in development" and that it is actively reviewing the examples flagged by Which? where summaries didn't match the content of reviews.

We fundamentally disagree with the premise of this investigation. - Tripadvisor statement

That distinction matters legally and reputationally: acknowledging the tool is still evolving doesn't change the fact that it already runs today on millions of property pages visited by tourists making booking decisions. Which? is calling on Tripadvisor to add clear warnings alongside AI-generated summaries and to introduce a fast mechanism for flagging discrepancies between a summary and the actual content of reviews.

What it means for Polish users

Polish tourists planning trips abroad often use Tripadvisor as their first source of information about hotels, and the growing popularity of automated review summaries means fewer and fewer people read individual reviews in full. The case points to a risk broader than one platform: any service with millions of user reviews that deploys AI to automatically summarize content can unwittingly average out and smooth over signals of real danger, especially when they involve rarer but serious incidents.

The case also feeds into a broader debate about platform accountability for AI-generated content, a topic gaining traction in the European Union as well, as further transparency obligations for high-risk AI systems come into force. For now, no Polish equivalent law directly regulates this type of summary, but the Tripadvisor case could become a reference point in discussions about what standards should apply to consumer review platforms operating in the Polish market.

Sources: Tripadvisor AI summaries give glowing reviews to dangerous hotels, consumer watchdog finds (euronews.com), Tripadvisor AI review summaries are potentially life-threatening (commstrader.com)