A 148-Hotel Study Found Your Website Has a 0.02 Correlation With Getting Recommended by AI
A study of 148 luxury hotels finds website and schema factors explain almost none of the variation in AI recommendation frequency — Forbes and Michelin ratings explain 54.7% of it instead.
Structured-data completeness on a hotel’s own website correlates with how often AI systems recommend that property at just 0.02 — statistically indistinguishable from no relationship at all. That’s the central finding from a study of 148 luxury hotels published on Hospitality Net September 9 by Andrew Paul, Managing Director of Americas Great Resorts, examining what actually drives recommendation frequency across ChatGPT, Google AI Mode, and Gemini. Website variables combined with market factors explained only 2.8% of the variation in how often a hotel got recommended. A model built on Forbes Travel Guide ratings and Michelin Keys, by contrast, explained 54.7% of the variation — by far the strongest signal Paul tested. The gap shows up directly in the recommendation counts: Forbes Five-Star hotels averaged 13.4 AI recommendation slots against 2.6 for unrated properties, and three-Michelin-Key hotels averaged 13.0 slots against 3.8 for unrated ones. Even having an llms.txt file — the emerging standard for signaling machine-readable site content to AI crawlers — made essentially no difference: 5.1 average recommendations with one, 5.7 without.
Paul’s conclusion is that AI hotel selection happens upstream of anything a website can control — driven by independent third-party credentials rather than a property’s own schema markup or accessibility optimizations — and he splits the real challenge into three separate problems: getting into the AI consideration set at all, making a site machine-readable, and generating repeat recommendations, only one of which a website actually addresses.
That upstream framing matches a harder number circulating the same week: Cvent’s analysis of 6,000 venue listings found 94% of hotels fail to appear in AI search results at all, with Diamond-tier listings 50% more likely to earn an AI citation than Basic-tier ones — a credentialing effect that echoes Paul’s Forbes/Michelin finding even outside luxury leisure travel. Separately, Hospitality Net has reported hotel marketers still lag OTAs in AI-discovery fluency — but per Paul’s data, that literacy gap may matter less than which third-party credentials a property can actually earn.
Source: Hospitality Net — The Hotel Website May Not Be Where AI Decides Which Hotels Matter Auto-generated brief — verified before publishing.