Why Hotel CMOs Keep Rearranging Deck Chairs on the Titanic

The Chairs Look Fantastic

Look at what a good independent luxury hotel is doing this quarter.

New photography. The property has never looked better. A brand refresh that took nine months and three rounds of stakeholder review. A website redesign with a scroll effect the agency is very proud of. A social calendar. An influencer trip. A loyalty relaunch. A partnership with a fragrance house. A media plan. A CRM migration that will finally, this time, unify the guest profile.

None of that is stupid. Some of it is excellent work.

And every item on that list has an organizational home. A budget, an owner, a deliverable, a metric, a deadline, a slide in the quarterly deck, a human being whose performance review contains it. Usually several of those at once.

That is not an accident. That is the system operating exactly as designed. An organization can manage what it can name, assign, fund and count. Hotel marketing has spent decades building instruments for these things, and the instruments work. The people using them are not lazy and they are not fools. They are doing the job the organization defined, measuring it with the tools the organization purchased, and reporting it to the people who asked for that report.

Meanwhile, a traveler in Chicago opens ChatGPT and types: best luxury hotel in Chicago for an anniversary.

Three hotels come back. Yours is not among them.

Or yours is among them and the description is wrong.

Nothing about that appears anywhere in the building.

The Guest Who Never Existed

Here is what makes this failure different from the failures hotel marketing already knows how to handle.

When a campaign underperforms, you get a number. When a rate strategy misses, you get a number. When a guest abandons a booking on the third screen, you get a record of it, and somebody builds a retargeting sequence around that record and presents the recovery rate at the next meeting.

Hotels have spent twenty years building instruments to observe losses that happen inside the funnel. This one happens before the funnel starts.

No impression. No session. No bounce. No abandoned booking. No lost-booking report. The traveler existed. The hotel simply never observed them.

You did not lose a booking. You lost the chance to compete for one, and losing that produces no first-party event at all.

If that accumulates, whatever it costs will be buried inside broader demand performance rather than identified as anything in particular. It will look like softness. Softness gets attributed to the market, the calendar, the comp set, the weather, the election, the airlift. It does not get traced back, because there is nothing to trace it back to.

So how much traveler consideration is moving to the layer you cannot see?

Phocuswright measured where US travelers go to research and select trip components. Between its late-2024 and second-half-2025 survey waves, search engines fell from 51 percent to 36 percent. Generative AI platforms rose from 6 percent to 15 percent.

Search lost 15 points. Generative AI gained 9. Fifteen percent is not a takeover, and it does not have to be. It only has to happen at the point where the traveler is deciding which hotels are even candidates, which is exactly where it is happening.

Now look at what happens inside that channel.

On a single day, July 29, 2026, we put ten traveler questions to ChatGPT, Google AI Mode and Gemini across six US luxury markets. 180 answers. Those answers contained 824 ranked hotel recommendations. Across all six markets the platforms named 152 luxury properties at least once. 23 of them captured half the recommendations. In the average market, five hotels take half. The top three take 41 percent.

The full study, with every question, capture protocol and exclusion, is published at The AGR Luxury Hotel AI Visibility Index.

Before anyone comforts himself that the machine simply picked the better hotels, one finding from that fieldwork settles it. Mandarin Oriental, Miami closed permanently on May 31, 2025. The building was demolished by controlled implosion on April 12, 2026. On July 29, 108 days after the implosion, ChatGPT recommended it three times and Google AI Mode recommended it twice, including in its answer for the top five luxury hotels in Miami.

A pile of rubble outranked working hotels. Meanwhile the same body of research found Forbes Five-Star and Michelin Key properties that appeared in zero answers.

This is not a merit ranking. Whatever these systems are weighing, the quality of the hotel plainly did not decide that answer. Something in the record did, and the record can be wrong.

Everything in a Hotel Has an Owner

Revenue has an owner. Distribution has an owner. Reputation management has an owner. The website has an owner. The CRM has an owner. Paid media, social, PR, photography, spa, F&B, engineering, the linens, the pool chemicals.

The minibar has an owner. If a bottle goes missing there is a person whose job includes finding out why.

Now ask a simple question at your next executive meeting.

Who owns whether the public factual record describing this hotel is correct?

Not the brand story. Not the positioning. The facts. The room count. The amenity that closed. The restaurant that changed concept in 2023. The page that still describes a renovation as upcoming. The two pages on your own site that contradict each other.

Ask it out loud and see what the answer sounds like. If it requires marketing to look at the web team, and the web team to look at whoever supplied the copy, and both of them to look at the agency, and the agency to explain that this is a client content matter, you have your answer. It is nobody.

Everyone can look at it and say, that's interesting.

And nobody has to touch it.

Which brings us to the money. A hotel can move $40,000 through marketing, procurement, legal, accounts payable and executive approval so that one person can publish one photograph. That transaction runs on rails. There is a name attached at every stage.

Correcting a factual error the hotel published about itself costs somebody an afternoon. The $40,000 has a process. The afternoon may not have an owner.

Now the fair objection. Much of that record is not yours to edit. In the July fieldwork, all ten of ChatGPT's Los Angeles answers cited two Michelin Guide pages. All ten of its Chicago answers cited two Tripadvisor pages. You cannot log into Michelin and fix your entry.

That is the harder half of the problem, and it makes the easier half matter more. Those pages get built from somewhere. When the sources describing your property disagree, you do not get a vote on which version survives. The minimum available to you is making sure everything you publish about your hotel is accurate and says the same thing everywhere.

Nobody owns that either.

Then Somebody Hands You the Report

Everything up to this point is an explanation. The loss is invisible. The instruments do not detect it. The org chart does not contain it. Nobody built this system on purpose and nobody inside it is doing anything unusual.

That explanation has an expiration date, and the date is the day the report lands on the desk.

Not a strategy deck. Not a point of view on the future of search. Not a vendor explaining that AI is changing everything and here are four trends. A document that says:

This page on your site says one thing. This page says another.

This amenity is described incorrectly.

This property fact stopped being true two years ago.

Here is the question a traveler asks. Here is the captured answer. Here are the three properties named. Yours is not one of them.

Here is a wrong fact about your hotel, appearing in the machine's answer, and appearing in your own published record.

Before that document, the blindness had an explanation.

After it, the blindness has a name on it.

Everything before the report describes an organization. Everything after it describes a decision.

Don't Believe Me

Now the obvious objection, which I will make for you.

At Americas Great Resorts, we sell the fix. Applying a discount to a diagnosis delivered by the man holding the cure is not paranoia, it is basic hygiene, and you should do it.

So do it properly. Take me out of the transaction entirely.

Open the two pages. They contradict each other or they don't.

Check the amenity. It is accurate or it isn't.

Type the question into the machine yourself. Ninety seconds, no cost, no vendor. Your hotel appears in the answer or it doesn't.

Look at the fact. It is current or it is obsolete.

You do not need to hire anyone to verify any of that. You need a browser and the willingness to look.

Because a commercial motive does not make a factual error false. Discrediting the messenger does not repair the information. The page stays broken either way.

And this is where it gets strange.

Nobody has ever written back to tell me a finding was wrong.

Not once. Nobody has said we checked item seven and you have it backwards. Nobody has challenged the capture, produced contrary evidence, or explained why the contradiction on their own site is fine. Nobody has said this is immaterial and here is the reasoning.

They go quiet.

I do not know what the quiet means, and I am not going to pretend I do. Maybe they think I am wrong. Maybe they think I am right and it does not matter. Maybe they think I am right and it matters and nobody knows whose job it is. Maybe somebody knows whose job it is and would prefer it not be his.

Silence preserves all four. That is what makes it such an efficient answer.

What silence does operationally is not ambiguous at all. The two pages still contradict each other. The obsolete fact is still published. Nothing the hotel actually controls has been corrected.

And once the report has been read, the silence is a decision. Nobody types DECISION: IGNORE into a document. Organizations make decisions this way constantly. This is one of them.

The Chairs Still Look Fantastic

The campaign shipped. The photographs are beautiful. The influencer got paid. The site launched on schedule. The quarterly deck says marketing delivered everything it committed to, and the deck is telling the truth.

Somewhere outside that meeting, a traveler asked a machine about your hotel and got a fact that was wrong. The same wrong fact is sitting on your own website right now, published by you, where anyone can find it. Including you.

You can no longer say nobody told you.

It was delivered in writing.

AGR will run an AI Visibility Audit on your property and show you exactly what the machines are saying about you and where the bad information lives. Free. No pitch. No obligation. Delivered as a PDF. All you have to do is ask.