This insight summarises Read before the guest: Booking.com's AI answers before you can, by Markus Busch, Editor/Publisher of hospitality.today, published on Hospitality Net on 26 October 2026.
The platform answers first. Busch examines pre-booking property answers generated from listings, photos and guest reviews. A traveller's question may be answered without anyone at the hotel seeing or composing the response.
Incomplete content leaves room for error. Outdated reviews or missing details can shape the answer a prospective guest receives. The article distinguishes these platform responses from the hotel's post-booking Auto-Reply messaging, where the property sets reply topics.
Visibility remains a question. Busch highlights uncertainty about what hotels can see and correct in these conversations. His discussion of Booking.com's rollout and Expedia's announced tools describes the position at publication; it should not be read as a current availability guide.
AIHA practical takeaway. Create a recurring content check around the questions guests ask reception. Test the property's available AI answers, then compare them with current operational facts and the listing. Record the date, market, question, answer and supporting evidence. Keep exact observations separate from assumptions about how the platform generated them.
Assign responsibility for checking facilities, room descriptions, policies and photographs when something changes at the hotel. Prioritise details that affect a booking decision or a guest's ability to use the property. Where an answer is wrong, update information under the hotel's control and use the platform's available reporting route. Recheck the result rather than assuming a content edit immediately changes every generated answer.
Why this matters for hotels. Content maintenance becomes part of the pre-arrival experience when intermediaries explain the property to travellers. A practical audit gives commercial and operations teams a shared record of what the hotel actually offers and what guests may be told. It also makes unresolved errors visible without promising control over the platform's model or customer conversations.