She's the only one who knows what the AI got wrong about your hotel

An AI assistant described your property before she booked. Your front desk is the one place anyone learns whether it was true, and nobody there is asking.

by Markus Busch

A guest walks up to your desk tonight carrying a description of your hotel you have never read. An AI assistant wrote it for her, in its own voice, out of listings, reviews and old press: the restaurant, the renovation, the quiet rooms at the back. She reads a hotel website with some suspicion. This she took as fact, because a machine said it flat, and she paid on it.

Some of it is right. Some of it describes the hotel you were two years ago, because the sources agree and they agree because they are old. Which parts are which gets settled in the next ninety seconds. If anybody asks.

The correction has nowhere to go

When a stay goes wrong, the complaint travels well. It reaches the platform, and the platform sends it to you. The whole business is wired to carry a problem toward the hotel.

Nothing carries a correction back. The AI assistant never learns that your dining room closed last spring. It wasn't at your desk and didn't see her face fall. The next traveler who asks gets the same answer, just as sure of itself. Booking.com hears about it only if she writes a review saying dinner was a letdown, and that review is about dinner. It says nothing about the sentence that sold her a restaurant you no longer run.

Your revenue manager never sees it. Neither does your market manager at the OTA.

So the gap between what she was told and what she found shows up in exactly one place: your lobby, at check-in. Then she takes the lift and it's gone.

What she knows expires by Thursday

Your post-stay survey and your reviews both arrive after checkout, and both ask how the stay went. Neither asks what she expected when she walked in.

She knows three things nobody else has. What she was told, who told her, and whether it held.

The AI assistant said the spa is open daily. It closes Tuesdays, and tomorrow is Tuesday. Or it said the courtyard rooms are quiet. They are, and that is why she picked you over the place by the station.

She won't put either in a review. By Thursday she couldn't repeat the sentence if you asked her to.

One question at the desk

Whoever is already checking her in asks it, while she still has her bag:

What had you heard about us before you came, and where did you read it?

It asks about a fact, so it doesn't sound like marketing. Guests answer it easily, and half the time the source is still open on their phone. It works for the same reason the email she gives you for a reason works: it comes from the person in front of her, attached to something real.

Then someone writes two lines in the shift log, or wherever your team already writes things down. What she was told. Whether it was true.

It takes ten seconds. No survey tool, no new system, no training beyond telling the team why you want the answer.

What a month of those lines buys you

Start with what it can't buy. You can't edit what an AI assistant says. You can fix the amenity line on your Google Business Profile, correct the OTA listing that still describes the old restaurant, rewrite the page selling last year's renovation. Do all of it. Nobody can tell you whether the AI assistant's answer changes, or when. If fixing the machine is the payoff you want, skip the question.

The payoff is on your side of the door, and it comes in two parts.

First, the desk stops getting ambushed. A wrong claim about the spa means the same disappointed conversation every Tuesday until the season ends, arriving as news to whoever is on shift. Learn it once, put it in the arrival briefing, and the desk says it first: the spa is closed tomorrow, shall I book you into the one on the square? Same fact, ten hours earlier. It lands as service instead of a complaint.

Second, and worth more: you learn which true thing about your hotel is closing guests. When she says she came for the quiet courtyard rooms, you have the reason a guest chose you without comparing you to anyone. Your booking engine can't tell you that. She never shopped.

Put that line everywhere your own words still travel intact. Your website, the OTA description, the GDS description a corporate booker reads. You can't make an AI assistant name your hotel. You can make sure the thing it repeats about you is the thing that has been selling rooms all year.

What it won't tell you

The answers come back uneven. Some guests remember what they read and some don't. Some want to talk, some want the key. On a Friday in August with four people in line, nobody is asking anything.

Each answer tells you one thing: a claim is out there, and one guest believed it enough to book. It can't tell you how many others heard the same thing. Read it as a tip-off.

That is still more than anyone else in the chain has. The platform sees a booking and a rate. The AI assistant sees a question it already answered. Your front desk sees her face when she walks in.

Tonight some of your arrivals are carrying a version of your hotel nobody on your staff has read.

Ask them before they go up.

Read also: She didn't choose your hotel. She accepted it.

Enjoying this analysis?* Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →