Wyndham Is Courting AI Agents. You Don't Have To.

On an earnings call on October 22, Wyndham's CEO started naming names. Meta's Muse. OpenAI's Dots. Instinct. He talked about them the way you'd talk about a new regional sales rep, someone worth courting because they bring business in the door. His argument was simple and, frankly, correct: an AI assistant that books a Wyndham room directly might cost the company less than a room booked through an OTA. Morgan Stanley puts OTA commissions somewhere between 10 and 20 percent. A direct booking runs closer to 5. So of course Wyndham wants the robots coming to its own website instead of shopping around on Booking or Expedia first. Wyndham also said its AI voice system has handled more than three million calls this year across over 1,700 hotels and lifted direct bookings by more than five points. That's not a press release line, that's a real shift in where the money lands.

Here's the part that matters to you. You were not on that call. Nobody from OpenAI or Meta is going to schedule a meeting with a four-table restaurant or a six-cabin outfitter to discuss integration terms. When Google rolled out hotel booking inside its AI Mode for U.S. travelers, the launch list had ten names on it. Big hotel groups, big platforms. Not one independent hotel. That wasn't an oversight. That's how every one of these rollouts works, and it will keep working that way, because it is easier for a product team to sign one deal with a hotel group that controls four thousand properties than to onboard four thousand individual owners one at a time.

So if your mental model of agentic booking is "get my business into the deal Wyndham just got," you can stop right there. That deal is not coming to you, this year or probably ever. The good news, and I mean this without irony, is that you don't need it.

Why a hotel room is easy to steal and your Tuesday menu isn't

Think about what an AI assistant is actually built to do when it books travel. It searches, it compares, it picks the best match against whatever the traveler asked for. That works beautifully when the thing being compared is fungible. A king room with a view of the parking garage at one Hilton is functionally the same king room at another Hilton three blocks over. The AI doesn't need a relationship with either property to do its job well. It just needs the price, the dates, and the cancellation policy. That is precisely the dynamic that has OTA executives nervous right now: an analysis of the sector this month pointed out that agents can line up identical hotel rooms across suppliers and just pick the cheapest one, which weakens the whole reason Booking and Expedia exist. If an agent can do the comparison shopping better than the OTA's own search bar, the OTA's moat starts to look thin.

The same analysis made a point that I think gets missed by almost everyone writing about this stuff: Airbnb's listings are harder to substitute, because a lot of them are specific, weird, one-of-one places. A houseboat on a particular lake. A converted grain silo outside Marfa. There's no equivalent two blocks over. You, the private chef, the dive guide, the sommelier running tastings out of your own cellar, are closer to that houseboat than to the Hilton king room. Nobody can find "the same thing cheaper," because there is no same thing. Your Friday tasting menu is built around what came off the boat that morning. Your guide's sunrise put-in on that particular bend of river isn't offered by the outfit across town, not really, even if they both call it a "river tour" on their websites. An assistant trying to arbitrage you on price is going to fail, structurally, because the comparison it needs doesn't exist.

That should be reassuring. It is also not the whole story.

The real risk isn't losing the booking. It's losing the conversation.

Skift reported something worth sitting with this month: Google's AI Mode can complete a hotel booking while leaving the hotel or the OTA as merchant of record, meaning the money still flows the way it used to, but the business on the receiving end may get almost no context about the conversation that produced the sale. No idea why the guest picked that property. No idea what they asked about. No idea what almost made them book somewhere else instead.

For a hotel chain running ten thousand rooms, that gap is a data problem, annoying but survivable. For you, it might be the whole business. When someone calls your restaurant at 9pm to book a table for an anniversary, half of what makes that booking worth anything isn't the reservation itself, it's the fifteen seconds where they mention it's twenty years, could you maybe do something with the dessert. You remember that. You build around it. A guide who's been running the same stretch of coastline for a decade can tell inside two minutes of a phone call whether a group wants white-knuckle current or a slow float with a cooler of beer, and that read is worth more to the trip than anything printed on the website. That's the actual product for a lot of independent operators. It was never just the activity. It was the fact that a person who knew what they were doing was listening to you describe what you wanted.

If an AI assistant becomes the layer that handles that exchange, summarizing "wants something scenic, moderate pace, two kids under ten" into a line item in a booking payload, you can still get the booking and lose the thing that made your business worth choosing over the next guide down the dock. You get a reservation. You don't get the person.

There's a second wrinkle worth naming plainly, because it changes how settled a "confirmed" booking actually is. Hilton has said publicly that it expects AI agents to keep re-shopping bookings after they're made, right up until the stay, checking for a better price or a better fit. Skift cited a study where half of a thousand U.S. travelers had changed a confirmed stay after finding something better. If you're a chef who buys eight pounds of halibut on Tuesday morning for a table of eight on Friday night, that re-shopping habit is not an abstraction. It's the difference between a deposit you can plan a shopping list around and one you can't. Nobody's built the etiquette for this yet. The platforms are still arguing about who holds the merchant relationship while this keeps happening underneath them.

What to actually do about it, since waiting isn't a strategy

You're not getting the Wyndham deal, so stop checking whether your booking software has signed a partnership with Muse or Dots or whatever assistant launches next quarter. That's not where your leverage is. Your leverage is in being legible on the plain, boring, open web, because that is where most agents are going to find you for a long time, launch partnerships or not.

That means the specific, non-fungible things about your business need to live somewhere an AI can actually read them in sentences, not buried in a PDF menu, not only in an Instagram caption, not locked behind a widget that needs five clicks to reveal a price. Your actual offering, your actual availability, your actual policies, written in plain text on your own site. If a guest relationship is going to survive an AI assistant acting as go-between, it starts with the agent being able to describe you accurately to the person it's working for. Right now, a lot of independent operators can't say with any confidence what an AI assistant would even tell someone who asked about them. That's worth finding out before you worry about which giant company is cutting which distribution deal this month. Bookaible has a way to check exactly that, how visible and how accurately described you already are to the systems doing this searching on a guest's behalf.

The chains are going to keep making noise about their deals, because that's the business they're in, managing supply at scale and squeezing margin out of distribution. You're in a different business. You're in the business of being the thing nobody else has, which is exactly what makes you hard for an algorithm to shop you away from. Don't spend your attention chasing a seat at a table you were never going to be invited to. Spend it making sure the thing you actually do is written down somewhere an agent can find and understand it. That's the whole job this month.