In 1998 we warned you the online travel agencies would end up owning your guests. Not renting you a booking. Owning the traveler, the discovery, the data, the relationship, while you supplied a room and a rate. You called it progress and went back to eating your lunch. Twenty-eight years later the discovery belongs to Expedia and Booking, and you rent your own travelers back from them every time. That warning is on the site if you want to see the date on it. It’s March 15, 1998.
It is running again, one floor up, and most of you are too busy eating your lunch to notice.
The OTA took discovery once the traveler was already shopping for a hotel. Artificial intelligence takes the moment before that, when the traveler asks a machine which hotels are even worth considering. The machine builds its answer out of whatever the public record says about you, before the traveler has typed a single hotel’s name. Get named and you get the shot at the traveler. Get skipped and you get none of that. No attribution, no traffic, no guest, no relationship. And here is the part that should bother you more than 1998 ever did. The OTA at least sold you a way back in. You paid the commission, you got the booking. The AI answer works nothing like that. You cannot buy your way into it. There is no commission, no ad slot, no invoice that puts your name there. You are in the record the machine reads, or you are out of the answer, and most hotels do not even know what that record says about them, let alone that it is the thing deciding whether they exist.
So the only question that matters is how you accurately get into the source record the machine reads. In 1998 all we had was a warning about the trap. This time we ran a test, on the hardest target we could find, to see whether there was a way out.
Best hotels in New York City. The most fought-over term in luxury hotel search, held for twenty years by Booking, Expedia, Condé Nast, and Forbes. We published one page aimed at that term, built to be read by the machine and nothing else. Three days later it was sitting in second place on Google for that query, above Condé Nast and above Forbes, names that have been ranking hotels since before the search bar existed. Then it got stranger. On July 29, when Google generated its AI answer for the ten best hotels in New York City, our page was the source cited on seven of the ten hotels it named. Aman, the St. Regis, The Carlyle, The Mark, the Ritz-Carlton Central Park, the Four Seasons Downtown, Casa Cipriani. None of them hired us. None of them knew we existed. The whole thing is documented on the site, every screenshot, including the mornings the ranking slipped and the day the machine cited us and got our own name wrong. Go read it if you do not believe it, because the rest of this depends on you believing it.
We are not going to stand here and tell you we know for a fact what put us there. We do not, and anyone who claims that kind of certainty about Google is selling you something. Here is what we will say. A page three days old, with no history and no backlinks, does not land on seven of ten judgments on the most fought-over hotel query in the country by accident. Age did not do it. The domain did not do it. Something formed that answer, out of the record Google reads. We build the thing that forms that record. So we will put the question to you the way we put it to ourselves, and then we will let you answer it: what else would explain it? We looked, and we could not find it.
The thing we build has a name. Knowledge Formation Optimization. It is work done on the public source record a machine reads before it answers, so that a property is described accurately and shows up where it should when the question gets asked. That is the whole of it, and the full version is on the site.
And that is the point, because it means the answer is not for sale but it is not sealed either. It is made of the record, and the record is something a hotel can write, before it hardens into whatever it is going to say. That is the opening 1998 never gave anyone. A chance to own the thing before the gatekeeper finishes closing the door.
That New York page carried ten hotels at once. Point the same machinery at one property, on the query that names its own city, and the position is not shared with nine other names. Here is what a position like that could be worth.
Take one luxury hotel, in its own city, showing up the way Aman showed up, on the version of the query a traveler actually types: best hotel in that city. Assume that position sends the hotel 1,200 click-throughs a month. We are choosing that number to run the math, not reading it off a live account, and you should hold it exactly that loosely. A click-through is a person who saw the answer, saw the property named, and clicked through to it. A click is a click. It is not a booking, and it is not proof anybody meant to book. We are not going to dress it up as more than a click.
Convert it at 1 to 2.5 percent. We will not call that band conservative, because published hotel conversion numbers run all over the place and some put five-star properties under 1 percent on a first visit. We chose it because a room at this price is a considered purchase, not an impulse, and considered purchases convert low on a first look. Take the bottom of the band if you think we are being generous.
At 1 percent that is 12 bookings a month. At 2.5 percent, 30. Put a booking at $2,400, which assumes a nightly rate near $800 across a three-night stay. That rate and that stay are assumptions too, not a verified average for any one market, so treat them as part of the model and not as a fact we handed you. On those inputs the model puts the position between $28,800 and $72,000 a month, or between $345,600 and $864,000 a year.
Now cut the traffic and the conversion both in half, because a hostile reader is allowed to attack both. You land between $86,400 and $216,000 a year. Still not a rounding error. That is the point of showing you the wrecked version. The case does not lean on the good end of the two assumptions a skeptic goes after first.
None of that is money the hotel is earning today. It is money the model says the position could be worth if the assumptions hold, and it is not revenue you already had, run through a cheaper channel. It is revenue that walks to whatever name the machine gives back. Some of the travelers who ask would have found you another way, and those you would have gotten regardless. The rest, the ones with no other path to you that day, book one of the names on the list, and if yours is not on it, that is the part you lose. Call the arithmetic above the ceiling on what being absent can cost you, not a bill you are handed all at once.
So here is a position, on the most fought-over term in the market, worth a real number even after you cut the traffic and the conversion both in half, sitting in an answer that none of the hotels in it worked for. An owner can see the screenshot. An owner can follow the arithmetic. And the owner will do nothing.
Not because the owner is stupid. Because the trade got turned around backwards.
The 1998 deal was easy to sign because the good part came first. The booking landed this week, and the bill, the slow surrender of your own demand, arrived over twenty years, quietly, on every folio. This is the same trade run backwards. Now the cost comes first and the payoff is out of sight, and the same reflex that said yes to the OTA says no to the fix. Fund the thing you can point to, skip the thing you cannot. Walk it through a real building and you can see why. Someone has to champion the spend and someone has to sign it. Often the champion is the commercial director or the head of marketing, and the signature is ownership or the asset manager. Look at what the champion is being handed: a number that will never appear next to their spend in any report, defended in a review that comes every quarter, on a payoff that arrives long after the review does. The rational move for that person is to wait and let some other hotel prove it first. That is not cowardice. It is a clean reading of their own incentives, and enough people in that seat read it the same way that almost nobody moves.
The one person in the building whose clock matches this trade is the owner. The commercial director is measured on the quarter. The asset manager is measured on the hold. The owner is the one still holding the building years from now, when the record that describes it has hardened into whatever it hardened into, and the one who collects for getting in before it did. And the owner is the person most likely to be told this is a marketing detail beneath their attention.
Here is the tell, and it is the same one from 1998. This building spends money on things it cannot cleanly attribute all the time. Brand. Public relations. The rebrand, the photographer, the sponsorship. Not one of those drops a tracked booking on the books the way a paid ad does. They get funded every year, because they are old and familiar and nobody ever got fired for approving them. The fear was never of spending on the hard-to-measure. The fear is of being the first name on a new one.
In 1998 we handed you a warning. You called it progress and went back to your lunch, and then you spent twenty-eight years renting your own travelers back from the companies that found them first.
This time we did not just warn you. We ran the test, and we handed you the result and the number a position like it could be worth. It is dated, it is on the site, and the answer is being written right now, out of whatever is lying around, exactly the way the OTA wrote the booking in 1998. You can work to put your name in that answer, or you can wait for it to harden and go back to filling rooms someone else sends you.
Most of you already know which one you are going to do. Twenty-eight years ago somebody handed you the warning. Now somebody has run the test, dated it, priced it, screenshotted it. You are going to call it interesting.
And go back to eating your lunch.
