A Hotel Demolished in April Is Still Showing Up in AI Travel Recommendations

Hotel CMOs are pouring budget into photography, website redesigns, and campaigns, according to Hospitality Net’s Andrew Paul, while a cheaper and more consequential problem goes unowned: nobody in most hotel organizations is responsible for making sure the property’s factual record — address, amenities, room counts, whether the building still exists — is accurate everywhere AI platforms pull from. The stakes are rising because search behavior is shifting fast: Phocuswright data cited in the piece shows search-engine usage among U.S. travelers fell from 51% to 36% between late 2024 and mid-2025, while generative-AI adoption for travel research rose from 6% to 15% over the same window. A study of 180 AI responses across six U.S. luxury markets generated 824 ranked hotel recommendations, with the top 23 properties capturing half of all slots — meaning AI-driven discovery is winner-take-most, and getting excluded from that top tier is costly in a way a stale webpage never was.

The clearest illustration is a hotel that no longer exists. The Mandarin Oriental Miami, demolished in an implosion on April 12, 2026, was still being actively recommended by AI systems 108 days later — an error first surfaced in an Americas Great Resorts audit published in early August, which found ChatGPT surfacing the demolished property three times and Google AI Mode twice, out of the same 824-recommendation dataset the newer piece cites. That the same phantom listing was documented publicly three weeks earlier and still hadn’t been corrected is Paul’s argument made concrete: unlike a $40,000 photo shoot with an owner and a budget line, fixing a wrong fact about a hotel “may not have an owner” at all. His prescription is organizational, not technical — assign clear responsibility for correcting contradictory information across owned and third-party sources, before AI platforms make that correction on the hotel’s behalf by excluding it.