Travel discovery is moving from search engines to AI assistants. This report analyses 16,295 hospitality videos across TikTok and Instagram to answer a single question: what does AI actually see when a creator talks about a hotel — and what does it miss?
Of the 2.83 billion views in this dataset, 2.39 billion (84.2%) came from videos where the words “hotel”, “resort”, or a brand name were spoken aloud but never written in the caption. AI models cannot read spoken audio. This is the structural reason AI gets hospitality recommendations wrong: the evidence it needs is sitting in 13,000 videos it cannot parse.
Six binary dimensions — each either present or absent across the video’s spoken words and caption combined. The result: only 15.1% of hospitality videos score high enough to be cited by an AI assistant. The remaining 84.9% exist in what we call the shadow reach — seen by humans, invisible to machines.
The single biggest lever? Caption completeness. Only 28.3% of hospitality videos have a caption of 50+ characters that mentions a hotel or location. This is the gap AI cannot bridge without human input.
Our data covered over 10 major hospitality brands, however here we focus on the AI legibility score vs. video volume of four key players. What we found? More videos does not mean more AI attribution. The data on the Four Seasons shows they lead on volume but score below peers because creators rarely name them in captions.