In AI search, the hotel's own website barely counts
AI assistants cite the hotel's own website as their source under 10% of the time. The recommendation is built from OTAs, review sites, and editorial guides instead.
Driving the news. New research from Tharro, published June 23, finds that when an AI assistant recommends a hotel, it almost never points to the hotel's own website as the source. Across 695 unbranded searches run through ChatGPT and Google's AI Mode, a hotel's own domain was the cited source under 10% of the time. The one channel a property fully controls is barely where AI learns what to recommend.
The finding. One distinction matters first. A mention is the hotel being named in the answer — the recommendation a traveler reads and acts on. A citation is the link the assistant footnotes as its source, and those links carry what little referral traffic AI sends. A hotel can be named without its site ever cited, and that is the common case. The gap holds even for hotels that have done the SEO work. When AI named a hotel that ranked its own website in Google, it cited that site as the source only about 20% of the time — and even for hotels sitting in Google's top 10, just 37%. The citations went somewhere else: to OTAs like Booking and Expedia, to review platforms like Tripadvisor and Oyster, and to editorial guides and regional "best hotels in X" roundups. Tharro's reading is blunt — AI learns which hotels to recommend by reading intermediaries, not hotels' own sites. (Tharro sells an AI-visibility tracking product, so the study doubles as a case for its own service; the citation figures are its own, from a single June 2026 snapshot.)
What it means for hotels. A decade of direct-booking strategy has been built on owning the relationship — the site, the booking engine, the guest data — and depending less on the intermediaries that charge to stand between hotel and traveler. AI discovery runs the other way. The owned site still serves the guest who clicks through and scrolls, but it is not what places the hotel in the AI answer. That visibility is assembled from sources the hotel does not control, and the sources AI reads most are the same OTAs and review platforms the direct-booking push was meant to lean on less. The mechanism is not mysterious: a hotel's own site speaks once, in its own voice, while the intermediaries hold dense, cross-referenced, frequently updated descriptions of the same property — the kind of corroborated material a model leans on. In the discovery layer taking shape, the hotel's own channel is a bystander to its own recommendation.
The catch. The study is one snapshot, not a verdict. Tharro is clear that its June 2026 data shows association, not cause, and the full study has not yet published. It is vendor research with a commercial stake, and the figures should be read as a directional signal rather than a settled measure.
The bottom line. In the AI discovery layer, the source is rarely the hotel. It is the intermediaries the hotel has spent years trying to need less.
Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →