An AI Looked for Your Hotel and Found Nothing
A researcher ran more than 27,000 queries against an AI system last month, testing it against over 650 hotels in Barcelona. The question was simple: does the AI know this place exists? For the big chains, almost always yes. For independents, the system came up blank eight times more often. Not ranked low. Not buried on page three. Blank, like the hotel never checked in to the internet at all.
That number should bother you more than anything else in the agentic booking conversation this month. Booking Holdings just launched an AI agent called Vayo that can plan a trip across multiple services and then hand you off to book. Google is letting people book hotels straight out of a chat window through something it calls AI Mode. Expedia is live inside Meta's assistant. Every one of these systems works the same basic way: it decides what to show you before you ever get a chance to make your case. If the AI doesn't know you exist, none of that distribution fight matters, because you were never in the room.
Why would an AI skip right over you?
Think about how a chain property gets itself known. Marriott has a brand name that shows up in training data a thousand different ways: press releases, loyalty program pages, corporate filings, news coverage of quarterly earnings, Wikipedia entries updated by people who don't even work there. A Residence Inn in Tulsa is mediocre and forgettable and the AI still knows exactly what it is, because the brand did the work of being known at scale, years before any of this software existed.
Your twelve-room inn did not get that treatment. Nobody wrote a Wikipedia page about your property. No analyst mentioned you on an earnings call. The only version of you that exists online is whatever you built yourself: a website, maybe a listing on Google Business, maybe a profile on Booking.com that a consultant set up for you in 2019 and nobody has touched since. If that material is thin, outdated, or inconsistent, an AI system trying to answer "find me a quiet place near the trailhead in Chamonix" has nothing to grab onto. It's not ignoring you. It genuinely doesn't know you're there.
This is the part that should sting, because it means the fight isn't really about AI at all. It's about the same thing it's always been about: whether your information online is complete, current, and consistent, or whether it's a patchwork held together by habit. AI just made the cost of neglecting that visible in a new way. Before, a sloppy website meant you lost some fraction of search traffic to a competitor with better SEO. Now it can mean you don't exist to the system deciding who gets recommended at all.
The llms.txt file won't save you
There's a fix making the rounds in operator forums right now: add a file called llms.txt to your website, a kind of README for AI crawlers, telling them what your business is and what you offer. It sounds like exactly the kind of clever, low-effort trick an independent operator would want. Ten minutes of work, future-proof your visibility, done.
The same Barcelona research tested this directly. Hotels that added an llms.txt file showed no measurable visibility advantage over those that didn't. None. The thing that did correlate with visibility was Google review volume. Not review score, necessarily. Volume. The sheer quantity of people who had reviewed the place on Google.
That result is almost insulting in how unglamorous it is. Nobody wants to hear that the lever is the same lever it's always been: get more people to leave you a review on Google, keep your listing current, make sure your hours and your phone number and your address match everywhere they appear. There's no file to upload that substitutes for that. There's no agentic trick that skips the step where real guests tell the internet, in volume, that you exist and that you're good. If you've been putting off asking satisfied guests to leave a review because it felt like busywork, this is the finding that should change your mind. It is, apparently, the actual infrastructure of AI visibility, dressed up as something people do for Yelp.
I'd go further. If a vendor calls you this quarter selling an "AI optimization package" built around some proprietary file format or a badge for your website, ask them to show you independent evidence it moves the needle, not vendor copy. The best evidence available right now says the lever is boring, slow, and has nothing to do with software you'd buy from them.
What actually makes you visible, and what happens after
So say you do the unglamorous work. Reviews pile up, your Google Business listing is accurate down to the parking situation, your website states plainly what kind of place you are and who it's for. You get found. An AI travel agent surfaces you in a shortlist for a guest planning a long weekend. Now what?
This is where the research gets genuinely uncertain, and anyone telling you they know how it plays out is guessing. Hilton's technology leadership has argued publicly that AI agents could push bookings toward suppliers directly, cutting into OTA share, because the agent's job is to find the guest the best option, not the option that pays the biggest commission. That's a hopeful read for an independent with a decent direct booking setup. But Hospitality Net's reporting this month raises the opposite worry: AI-driven discovery can send a visitor to your website, get them interested, and then lose them to an OTA anyway, because the agent, or the guest, finishes the transaction somewhere with a familiar checkout flow. You did the work of getting discovered and still paid the toll.
A hotel distribution study from HOTREC this year found that 51% of surveyed hotels reported OTAs undercutting their own supplied rates, frequently or occasionally, which is its own old grievance that predates any of this AI conversation. The same research found hotels that had invested in their own booking experience saw a 51% direct booking share, which is a genuinely good number, and suggests the fundamentals of direct bookings still matter more than whatever AI layer sits on top. Meanwhile a separate survey found 70% of hotel chains expect AI-driven channels to increase their direct share over time. Optimism and anxiety are both well represented in the data, which tells you the honest answer is that nobody has settled this yet.
What's not in dispute is how small the current volume is. A Bernstein analyst told an audience at Skift's Data + AI Summit Europe this month that AI referrals remain below 2% of traffic at both Booking.com and Marriott, despite growing fast. So the visibility question and the attribution question are both live, but neither has resolved into a crisis yet. You have time. Not forever, but some.
There's also a trust problem sitting underneath all of this that nobody's solved. Kayak and Skyscanner executives, speaking at that same summit, pointed to price verification and plain trust as the real barrier to agentic booking taking off, not technical capability. A survey cited at the event found 51% of travelers were concerned about AI accuracy. An assistant that confidently books the wrong room type, or quotes a rate that's since changed, breaks trust fast, and that's a problem the platforms have to solve before volume grows much. It's one more reason the current moment rewards patience over panic.
There's a specific kind of damage worth naming here too: the zero-click answer. Some AI systems now answer a travel question directly in the chat window, without sending the user to a website at all. Good for the guest, maybe. Bad for you, because it erases the trail that used to tell you how someone found your place. No referral link, no "came from Google" tag in your analytics, nothing. You might get booked by an AI assistant and never know that's what happened. That's not a reason to despair. It's a reason to stop treating your website analytics as the whole truth of where your guests come from, and to pay closer attention to the raw count of direct inquiries and the reviews piling up on your Google listing instead, since that may be the signal that matters now whether or not you can see it in a dashboard.
If you want to know roughly how visible you actually are to the systems doing this discovery right now, bookaible has a simple way to check it, and it's worth five minutes before you spend money on anything fancier.
None of this requires you to reinvent your business. It requires you to treat your online presence the way you'd treat your dining room: clean, current, consistent, the kind of place that looks cared for the moment someone walks in, whether that someone has a pulse or is running on a few billion parameters deciding where to send a stranger for the weekend.