Question
Which sources do AI assistants cite when asked about hotels?
Short answer
Mostly intermediaries. Online travel agencies took 55.3% of citations in an audit of 1,357 Gemini citations, but the share flips by question type: non-OTA sources take 55.9% of experiential queries.
The one published audit of this looked at 1,357 grounding citations across 156 paired hotel queries in Tokyo. Across everything, online travel agencies held 55.3% of citations, which is the number most people quote.
The more useful finding is underneath it. Split the queries by what the traveller is trying to do and the picture inverts. On experiential questions, what a neighbourhood is like, what a property is near, what kind of stay it is, non-OTA sources took 55.9% of citations. On transactional questions, about booking and price, they took 30.8%. A gap of 25.1 points.
For a hotel or a destination that is the whole strategy in one line. Competing for citation on booking questions means competing with the intermediaries on their own ground, which is expensive and rarely winnable. Competing on experiential questions means describing the place accurately and specifically, which is work you control.
Why credible sources differ
One city, one model, one month, and a preprint that has not been peer-reviewed. The OTA classification rests on curated domain matching. Treat the direction as informative and the exact percentages as provisional until someone replicates it elsewhere.