In 2027, Your Hotel Has Two Guests. You're Probably Ignoring Both.
As AI agents begin driving hotel bookings, independent hotels must optimize machine-readable data for discoverability while redirecting AI-saved staff time toward personalized guest recognition.
Photo by Are Morch, AI Champion
AI will book the room. It cannot make anyone feel known. A vision for independent hotels in 2027.
Last week in New York, one of the most respected hotel owners in America stood on a Skift stage and said he wants to ban AI agents from booking his hotels.
His reason was honest. Agents make rational decisions. They don't buy on impulse, so they don't buy the upgrade, the late checkout, or the bottle of wine. Then he admitted the part that should keep every hotelier awake: he can't figure out how to stop them.
That is the 2027 problem in two sentences.
We have spent two years asking whether AI will change hotel bookings. Wrong question. The better one is this: who is your guest now?
My answer: starting in 2027, every hotel serves two guests. One of them never checks in.
I call this the Two-Guest Hotel. One guest is a machine that reads you and decides, in milliseconds, whether you are worth recommending. The other is a human who walks in hoping to be known. Your hotel now has to win both, and they ask opposite things of you. Miss the first and the second never books. Serving only the first and the second feels like a transaction. And the clock is not theoretical. Since September 30, a hotel that does not send Google its own rates can watch its price disappear from its own hotel card, replaced by whichever reseller does. The new channel is being wired right now, while it is still small, which is exactly when its rules get written.
Guest One: the machine that books
Guest One doesn't care about your lobby. It has never smelled your coffee or met your night auditor. It reads.
It reads your price feed, your Google profile, your policies, your reviews, and what every OTA and travel writer has ever said about you. Then it decides, in milliseconds, whether you are worth recommending.
Here is the uncomfortable part. When we studied 27,360 AI travel answers in Barcelona, hotel websites supplied only 7% of what AI said about a property. The other 93% came from OTAs, review platforms, and media. Independent hotels were eight times more likely than chains to get no mention at all.
So the story AI tells about your hotel is mostly written by someone else.
Guest One also never stops shopping. Hilton expects agents to keep reshopping a booking until departure, and half of U.S. travelers in a McKinsey and Skift study have already changed a confirmed stay after finding a better option. An agent will do that tirelessly, on every booking, forever.
Guest One is not coming. It is already reading you.
Guest One is getting a front door.
In August, Google started taking hotel bookings directly inside AI Mode. Ten companies were invited at launch: the big OTAs and the big chains. Not one independent hotel.
That is the headline most people remember. The more important story is what came next.
Google published an open standard (UCP) and says it will use it to power hotel booking in AI Mode. The hotel stays the merchant of record and keeps the guest relationship.
Meta's Muse agent is getting a direct line to independent hotels through at least one PMS, with bookings landing as direct reservations. The U.S. rollout starts in Q1 2027. (Full transparency: that PMS is Cloudbeds, and I'm a Cloudbeds Ambassador.)
On November 10, OpenAI opened profile claiming in ChatGPT, so a hotel can correct what ChatGPT says about it and see how people engage.
The doors exist. They are not locked. But nobody is going to carry an independent hotel through them.
Let me be honest about the numbers, too. Booking Holdings says AI assistants still bring in well under 1% of its room nights. Guest One is small today. That is exactly why 2027 matters: the rules of a new channel are written while it is small, not after it is big.
Guest Two: the human who arrives
Then the machine sends you a person. Tired from the flight. Hoping the room is quiet. Wondering if anyone will remember she was here last spring.
This is where the AI story has quietly gone off course.
A new AHLA and Actabl report found that 84% of hotel respondents say AI has improved their own work. Only 15% say it has improved frontline work. Read that again. We have used AI to make managers faster. We have not yet used it to make guests feel more welcome.
Here is the paradox I think will define 2027. The more the booking gets automated, the more everything becomes comparable: price, points, perks. An agent can compare a discount. It cannot compare the moment your front desk greets a returning guest by name before she says a word. It cannot be the night auditor who remembers she likes a high floor away from the elevator and has quietly moved her there before she thinks to ask.
Skift Research put it well: on-property recognition is the part of loyalty an algorithm cannot list.
That is not a soft idea. That is the moat.
The trap: building in the wrong order
Most hotels I talk to are making one of two mistakes.
The first is buying guest-facing AI before fixing what Guest One reads. A beautiful chatbot on a website that supplies 7% of the story, with no price feed behind it, is a nice receptionist in a building nobody can find.
The second is using AI to cut costs, then wondering why Guest Two feels nothing special. If AI gave your team back ten hours this week and those hours disappeared into a smaller schedule, the promise of AI in hospitality was not kept.
Stephen Covey taught us to seek first to understand, then to be understood. In 2027, that principle has two directions:
For Guest One: make your hotel easy to understand, so the machine can recommend it with confidence.
For Guest Two: take the time AI gives back and spend it understanding your guest.
That is the sequence. Make the hotel legible before you make it intelligent. Make it human before you make it efficient.
The Two-Guest Hotel: five moves for 2027
None of these requires a big budget. All of them require a decision.
Feed. Send Google your own rates through Hotel Center or your booking engine's connectivity partner. If your price isn't on your own card, Guest One is quoting someone else's.
Claim. Own your Google Business Profile and claim your ChatGPT profile. Fix the story at the source instead of hoping the internet gets it right.
Connect. Ask your booking engine and PMS one question in writing: what is your plan for AI agents and Google's UCP, and when? Their answer tells you whether you are on the guest list for 2027.
Decide. Write a simple agent policy. Which AI agents may book you, through which door, at which rate? You can't ban the machines. You can choose how they come in.
Redeploy. Measure where the hours AI saves actually go. Put them on the frontline, then track it. If that 15% number doesn't move in your hotel, the technology is working for you but not for your guest.
Considered by the machine. Remembered by people.
The chains will outspend independents on AI in 2027. That race was never yours to win.
Your race is different. The machine will decide whether your hotel is considered. Only your people decide whether it is remembered. Independents don't need to out-spend the chains. They need to out-host them, and make sure the machine can find the door.
The owner on that Skift stage was right to be afraid and wrong about the fix. You cannot ban the machines. You can decide how they come in, and then make the welcome on the other side of that door so human that being considered turns into being remembered.
So here is my question for you as the year closes: which of your two guests is your hotel built for today? If you hesitated, you already know where 2027 starts.
If you want to find out where your hotel stands with both guests, walk through the questions I use in the AIDURIX Hotel AI Diagnosis: Where are we? What have we learned? What happens next?
The compass is ready. The direction is yours.
Comments
Comments for this content
0 comments available