As of June 2026, AI tools — ChatGPT, Gemini, Claude — became the most relied-upon travel planning resource, ahead of traditional travel websites for the first time: 58% versus 55%. This isn't early-adopter behaviour. Phocuswright tracked US traveler AI adoption for trip planning at 24% in 2024, 43% by late 2025, and 56% by March 2026 — mainstream arrival, not a niche trend. Travel and hospitality is the #1 industry by ChatGPT usage in the purchase journey, at 47% of buyers.
25 London boutique hotels, 9,380 AI answers
A study of 25 London boutique hotels found 4 hotels held 64.3% of all AI mentions, 12 hotels held under 1% share each, and 2 hotels did not exist to AI at all. This wasn't about quality — several of the invisible hotels were excellent. The AI simply couldn't read them. Separately, 94% of hotel websites were found to lack the technical and content foundations that let AI platforms read, understand, and cite the property.
AI reads facts, not poetry
"A stunning boutique retreat with world-class charm and unforgettable hospitality" extracts as nothing to an AI model. "11-room family-run hotel, 3rd generation since 1974, 200m from the beach, homemade breakfast from our own olive grove" extracts as five citable facts. Most hotel website copy is the first kind.
Booking.com describes you hundreds of times. You describe yourself once.
A hotel's listing is often repeated, word-for-word, across hundreds of OTA affiliate domains. AI reads that repetition as consensus — and can treat the hotel's own website as the outlier version of itself. Since ChatGPT integrated Booking.com and Expedia, AI often credits the OTA directly rather than the property, and every booking routed that way costs 15–25% commission — more for villa rentals through Airbnb or Vrbo.
Boutique and independent properties have an opening
Boutique, unique, and locally-recommended queries are disproportionately common in AI travel prompts, and independents can outperform chains on exactly those queries. Smaller, lower-ranked properties gain the most from optimization — chains are currently investing their AI budget in revenue management and guest messaging, not visibility. That's a door left open. A restaurant or hotel will never out-cite a global chain on volume, but it can absolutely own its postcode's answers.
The same structural fixes — schema, extractable copy, consistent entity information — are covered for restaurants specifically in 83% of Restaurants Are Invisible to AI, and applied end to end in the Arros QD case study.