Less Than 1% of Bookings Come From AI. For Now.

Dana runs a six-room farmhouse inn outside Phoenicia, New York. Restored barn, a trout stream on the property line, breakfast she cooks herself most mornings because she likes the quiet before guests come down. On a Tuesday in early October, a friend forwarded her a trade newsletter clip about Hilton, Marriott, and Hyatt warning investors that AI agents could divert bookings away from their own sites. The subject line was something like "this is coming for all of us." Dana read it twice, poured a second coffee, and spent the rest of the week trying to figure out if she should be scared.

That week is worth walking through, because it is probably close to what a lot of independent operators are doing right now. Not building anything. Not panicking out loud. Just quietly trying to figure out whether the ground is actually moving under them, or whether someone is selling fear to people who already have enough to worry about.

Is anyone actually getting booked by AI yet?

Dana's first move, Tuesday night, was to open ChatGPT and try to book her own inn through it, the way a guest might. Nothing happened. No reservation, no availability check, not even a confident wrong answer. The assistant suggested she call the inn directly or check a booking site. That's not a failure of the technology. It's accurate. As of this fall, agentic booking for a six-room independent inn basically does not exist as a live, working path.

The bigger platforms tell a similar story, just at bigger scale. Booking.com has reported that traffic from AI assistants sits below one percent of its room nights, and that figure hasn't moved much in recent quarters, according to trade reporting in early October. Google's AI Mode can now search, compare, and complete bookings for a limited set of hotel partners in the US, including Booking.com and Expedia properties, with the actual payment still processed by the hotel or OTA. Meta's Muse assistant is expected to carry Expedia inventory soon, though that integration wasn't confirmed live as of the most recent reporting. A newer entrant, Reservations Group, launched a conversational app at the end of September that claims to handle hotel, flight, and rental car bookings by text or voice, and says its systems already process a quarter million customer interactions a day. That's a company claim about interactions, not a verified count of completed bookings, and it's worth holding those two things apart.

For vacation rentals specifically, one industry analysis covering the Tampa Bay market put it plainly: no AI agent is booking rentals at scale there today. For restaurants, guides, and independent experience hosts, the picture is even thinner. The visible reporting is almost entirely about hotel rooms, flights, and cars. Nobody has published numbers on AI assistant bookings for a private chef's tasting menu or a half-day fishing charter, because as far as the public record shows, that volume barely exists yet.

So Dana's experiment wasn't a fluke. The honest answer to "is this happening to me right now" is: not yet, not really, and not in a way anyone can measure with confidence.

Why one percent still worries the big chains

Here's the part that should actually hold your attention, though. If AI assistant bookings are under one percent and barely moving, why are Hilton, Marriott, and Hyatt putting this in their annual reports as a named risk to investors?

Because annual reports aren't about this quarter. They're about the company's exposure if the number stops being one percent. A hotel chain with a global loyalty program and decades of direct-booking investment doesn't wait for the trend to become obvious before it tells shareholders about the risk. It flags the possibility of a new intermediary sitting between the guest and the brand, one that could route demand, shape comparisons, and quietly take a cut of a relationship the chain spent years building. That's not fear about today's traffic. It's fear about who controls tomorrow's shelf space.

For a small operator, the lesson isn't "panic now because Hilton is panicking." It's the opposite, in a way. Hilton has the resources to build direct-booking technology, loyalty data, and app integrations that can compete for relevance against any new agent that shows up. Dana doesn't. What she has is something the chains don't: a business small enough that an honest, well-structured page about her inn is basically all the infrastructure she needs.

There's a structural detail buried in recent industry research that makes this concrete. A survey of 343 property management system vendors found that only seven percent publish fully self-serve API documentation, the kind of technical doorway an AI agent would need to read live availability and book directly into a hotel's system. That's not a small-operator problem. That's an industry-wide one. Most of hospitality, big and small, currently has no clean door for an agent to walk through. The giants are racing to build theirs. Independent operators have a rare window where the infrastructure gap is roughly the same size for everyone.

What a six-room inn can do about it this week

By Thursday, Dana had talked to a friend who guides fly-fishing trips two towns over, and to the manager at a restaurant in town who'd just dealt with a cluster of suspicious reservation requests that looked machine-generated. Neither of them had a crisis story about losing bookings to an AI agent. What they had, all three of them, was the same quiet unease: not knowing whether their information online was even readable by whatever shows up next.

That's the actual, useful question for this week. Not "how do I get booked by AI," because nobody has a reliable playbook for that yet, including the OTAs spending billions on it. The useful question is: if an assistant tried to describe my business accurately right now, using only what's publicly available, would it get the basics right?

Check the obvious things first. Hours, actual ones, not the ones from 2019 that are still sitting on a map listing somewhere. Cancellation policy, written in plain sentences, not buried in a PDF confirmation email a guest only sees after booking. Dietary accommodation and allergy handling, if you're a chef or a restaurant, stated somewhere a machine could find it, not just known by the host who answers the phone. Photos that match what a guest will actually see when the car pulls up. Pricing that doesn't require a phone call to confirm. None of this is new advice. It's the same advice good hosts have always followed for human guests who do their homework before they show up. The difference now is that a research assistant doing that homework reads differently than a person does. It wants structure. It wants consistency across the places it looks. It does not forgive a phone number that's wrong on one site and right on another.

One data point worth sitting with: a recent hospitality study found that around fifty-six percent of engaged hotel website visitors show real booking intent, but only about two percent convert. That gap, between intent and action, is where confusion and friction live. Wrong information, slow pages, unclear policies. An AI agent reading your site on a guest's behalf will hit that same friction. If anything, it will be less patient with it than a human, because it has no reason to call you and ask a clarifying question. It will just move to the next option that's easier to read.

This is also where the OTA conversation gets more honest. One analysis of Booking Holdings and Expedia's divergent bets on AI shopping agents made a point worth taking seriously: independent properties often lack the direct-booking infrastructure that an agent could plug into, which means agents may end up routing demand through OTAs rather than around them. That's not the giants disappearing. It's the giants possibly becoming more necessary, not less, if independent operators don't put in the basic work of making their own information legible outside those channels. Direct bookings don't get easier just because an assistant exists. They get easier because the information an assistant needs is sitting in the open, accurate, and consistent, whether the booking happens on your own site, through an OTA, or eventually through whatever agent shows up next.

By Friday, Dana hadn't bought any software, hadn't hired anyone, and hadn't rebuilt her website. She'd spent two hours fixing her hours listing in four different places, rewriting her cancellation policy in plain English, and replacing three photos that no longer matched the rooms. Small, unglamorous, the kind of task that doesn't feel like doing anything about artificial intelligence at all. That's probably correct. The AI assistant bookings aren't here yet in any volume anyone can prove. The chains are worried about a future that hasn't fully arrived. What an independent operator can actually control this week isn't whether an agent books her inn. It's whether, when one finally tries, it finds something true.

If you want a plain read on how visible your own listings and information already are to the assistants poking around out there, that's the kind of check bookaible is built to give you. Not a prediction about next year. Just an honest look at what's findable right now.