The most valuable conversation in hotel sales used to happen before the booking. A guest asked if the parking was on site, or if the street-side rooms were quiet, and someone at reception gave a good answer. On Booking.com, that conversation now happens without you. And I think most hotels are still writing their listing as if nobody important reads it.
Somebody does. It is an AI, and it is answering your guests for you.
What is actually happening
Markus Busch has a sharp piece on Hospitality.today this week that sets out the facts plainly. Booking.com announced in October 2024 that travellers could ask specific questions about a property and get an instant answer. That answer is generated from three things: the property listing, guest reviews and photos. Nobody at the hotel writes it.
At launch it was live in the app in the US, the UK, Australia, New Zealand and Singapore, with property pages and local-language versions in Spain, Italy, Germany, France, Poland and the Netherlands promised later. Expedia showed its own versions, AI Property Expert and AI Property Compare, at its partner conference in May, though they have not launched and there is no date.
Google does something similar. Its place summaries are written from user reviews and labelled “Summarized with Gemini”. The only fix Busch found in Google’s documentation is a button to flag inappropriate content. There is no field for the business to correct the summary.
Hospitality Net’s daily brief put a cost on getting this wrong. It points to Aven Hospitality research this month that put the average cost of unmet expectations, driven by inaccurate content, at $1,900 per trip.
Your listing is now your sales team
Here is my view. This is not an OTA problem. It is a content problem that hotels have been ignoring for years, and AI has just made it visible.
For a long time, an Extranet listing was admin. Tick the facilities boxes, upload some photos, write a description once and forget it. The real selling happened on your own website, or when a guest emailed and a good receptionist won them over.
That model is gone for a growing share of guests. If your listing never says the garage is under the building, the AI cannot say it either. If three guests complained about a nightclub that closed in the spring, that complaint can still shape the answer about noise. Every blank field becomes a gap that strangers’ reviews will fill. Every old review becomes a salesperson you cannot retrain.
Look at the timing too. Booking.com gave hotels Auto-Reply in October 2025, so you can answer guest questions in your own words. But as Busch notes, that runs on guest messaging, which opens with the reservation. Your words mostly reach people who have already booked. The question that decides whether they book you or the hotel next door goes to Booking.com’s AI.
This is the same pattern I keep seeing across AI search. When I wrote about why AI hasn’t heard of many independent hotels, the cause was thin, scattered information. When Google wrote the rules for AI hotel booking, hotels were not in the room. Now the platforms that sell your rooms are also the ones describing them.
The fair counter-argument
Some will say this is overblown. Booking.com has not published how many travellers use the feature, or how often an AI answer comes before a booking. It has not even said where it runs today, two years after the announcement. Fair enough. We genuinely do not know the scale.
But that cuts both ways. If you do not know how often the AI is answering for you, you also do not know how often it is getting you wrong. And neither Booking.com nor Expedia has said whether hotels can see the questions guests ask, or correct the answers. I find that the bigger issue.
Those questions are some of the best demand data in travel. They tell you exactly what stops people booking: parking, cribs, late check-in, noise. Right now the platform keeps them. Hotels should be asking for that data loudly, at every partner meeting. As I argued when Booking started buying the trip layer, the platforms are building the conversation. Suppliers need to make sure they are not just the inventory inside it.
What I would do this week
Start with Busch’s simple test. If your hotel is in a market where the feature runs, open the Booking.com app, find your property and ask it the five questions your reception hears most. Write down the answers and compare them with your listing. Where an answer is wrong or missing, the listing is usually where it went wrong, and the listing is the one source you control.
Then treat your listings as owned media, not admin. Give someone the job of keeping every field complete and current across Booking.com, Expedia and Google. Fix the issues your reviews keep repeating, because those themes are now raw material for the answers guests see.
So here is the debate. Is the OTA listing still a distribution chore, or is it now the most important sales copy a hotel writes? I think it is the second, and the hotels that act first will win the bookings that are being decided before anyone at reception sees a question. Where do you stand?
Want more takes like this? I share what AI is really doing to travel search, booking and marketing every week. Subscribe to my LinkedIn newsletter.