Tripadvisor's AI summary called a hotel facing a lawsuit "spotless" while burying harassment complaints

The British consumer rights organisation Which? published an investigation on 2 July 2026 finding that Tripadvisor's AI review-summarisation feature systematically downplays serious guest complaints, up to and including sexual harassment and mass food poisoning.

What the algorithm hides

Which? researchers checked the AI summaries of several hotels and found a systemic pattern: the algorithm does not merely smooth over negative feedback. In some cases it produces a description that directly contradicts what guests actually experienced.

• A hotel facing a lawsuit over mass food poisoning was described by the algorithm as "spotless"
• A resort where guests widely complained of sexual harassment by staff received an AI summary reading "friendly service"
• In other cases, the algorithm said nothing about complaints of a mouldy smell and the lack of a central water supply
• Tripadvisor introduced automatic summarisation to aggregate millions of reviews worldwide

The gap between the AI summary and the actual reviews is not a technical inaccuracy like a slightly wrong star rating. It is a qualitative inversion: "spotless" and "lawsuit over mass poisoning" are mutually exclusive descriptions of the same place. The algorithm did not get a detail wrong; it reached a false conclusion.

Why algorithmic neutrality is dangerous

Travellers increasingly base booking decisions on the AI summary at the top of the page rather than on the full body of reviews. The convenience is obvious: no need to scroll through hundreds of comments spanning several years. But if the algorithm systematically plays down negatives, users lose information that is critical to their own safety.

An algorithm trained on millions of reviews is most likely optimised for "balance", leaning neither toward criticism nor toward praise. That kind of neutrality is acceptable when the disagreement is about taste: some people like the breakfast, others don't. But when complaints concern health and safety (food poisoning, harassment, unsanitary conditions), algorithmic neutrality turns into disinformation.

"Travellers should be able to trust what they read,"

Which? said in a statement following the investigation.

Platform responsibility is a separate question. When a human editor publishes misleading material, accountability is clear. When the author is an AI launched by the platform itself, the legal picture is murkier. The Which? investigation indirectly raises exactly this question: is Tripadvisor liable for an AI summary that could put users at risk?

What this means

The Which? investigation is another signal to the industry: automated summarisation of user-generated content needs a special regime for safety-related complaints. Reviews about poisoning and harassment are not just "negative opinions" among others; they are warnings that the algorithm must bring to the foreground, not average out. For Tripadvisor, this is a matter not only of reputation but also of trust in the platform as a tool for real consumer decisions.

Frequently asked questions

How does AI summarisation on Tripadvisor work?

The algorithm analyses thousands of user reviews of a hotel and generates a short text summary, which appears at the top of the property's page. The feature is meant to help users who have no time to read all the comments. According to the Which? investigation, it is precisely this summary that hides critically important complaints.

Who conducted the investigation?

Which? is an independent British non-profit consumer rights organisation founded in 1957. It regularly tests digital products and publishes warnings about consumer risks, including in the area of AI services.