Which travel apps train AI on your data?
Of the 6 travel apps on file, 3 are graded, and every one of them trains on your data by default. The other 3 have a policy that doesn't say either way. All 3 graded entries are D. Airbnb scores lowest for exposure and Uber highest.
Do they train on where you went, where you stayed and how you got there?
A travel app holds an itinerary of your life: the cities, the dates, the hotel, who was on the booking and what it cost, and for a navigation app the routes you drive. The question here is narrower than who that record is sold to. It’s whether it trains a model, and whether the document offers a way to keep it out. Read together, the travel apps divide cleanly. The ones that say so and name a way out, and the ones whose documents never reach the question. Some describe their AI at length without saying what it learns from. One never mentions it at all.
The ones that say enough to grade
Airbnb states the use and writes out a control. If you don’t want it to use your personal information to “develop and improve AI models that power certain features on Airbnb”, you can set your preferences “as described here”. The link leaves the policy, so the document never says where the preference sits or which way it ships. That’s the gap between a D and a C here. A help article, archived on the verdict, gives the route: Account settings, then Privacy, then Help improve AI-powered features, where the tick means on.
Booking.com’s clause is unqualified: it may use “your personal data” to “develop and enhance our machine learning models and artificial intelligence systems”. No category excluded, no product named. For a travel platform, that’s where you went, with whom and what you paid. The way out is a written objection, “subject to applicable law”, with no law named. A setting for AI Memories reaches one feature, not the models. The objection reaches the models, but it’s a letter, not a switch.
The ones that never say whether anything is trained on you
Expedia’s statement has a section headed “Our use of artificial intelligence”, with a table of purposes. It opens with pricing and fraud detection, runs through feature generation “To enrich other applications such as embeddings” and moderation of listings and reviews, and ends in a row for other uses. That’s data going into AI systems. Whether any model is trained on it isn’t said. Training appears in the document only as staff training, in call-recording clauses and in the recurring purposes phrase about market research and analytics, and none of it is a statement about models. Chats with the AI assistant are collected as communications. The section describes uses, not training, and that is the finding.
Hotels.com’s statement carries the same heading and names its AI uses one by one, pricing, fraud, chatbot, recommendations and images, before a final row headed “Other”, for uses “not captured by the above categories”. This isn’t silence. Having set out that much, it never says whether your data builds the models behind any of it. The phrase that recurs through its purposes table, “Market research, analytics and training purposes to improve our services”, neither says model training nor rules it out.
Waze’s policy never uses train, model, machine learning or artificial intelligence. Your information goes “to improve the Services or to develop new features or services”, and route information to improve “the accuracy of its mapping and navigation data”. As “a Google-owned entity” it shares de-identified route data with Google. Sharing isn’t training, and improving a map from routes isn’t training a model. The policy never says which it means. It names Privacy Settings and a right to object, but for a purpose it never states there’s nothing to switch off.
Side by side
2 pairs of travel apps people choose between, with both verdicts on one page.