Is Your Hotel Legible to AI? Start With the Facts, Not the Tool.

Marriott's CEO said it last week without hedging: the biggest impact AI will have on our industry is distribution — and Marriott is already feeding its content to Google and OpenAI so the machines have it first.

IHG's commercial chief followed with the line I keep coming back to: "AI tools can only be as helpful as the content they're drawing from." IHG is building a whole platform just to make its hotels legible to AI.

The chains have spotted the real bottleneck. It isn't the model. It's whether your hotel's facts exist somewhere a machine can read.

The demand is real and it's fast

• +800%. In China, hotel bookings made through phone AI assistants grew eightfold year-on-year at Spring Festival — the traveler never opened the booking app.
• 40% of travelers would fully delegate the comparison of options to an AI agent (Accenture, 5,003 travelers, 12 markets, June 2026).
• From 1–2% to 14–16% conversion. That's what one voice-AI reservations platform reports at 500,000 calls a day — human-level conversion, at machine scale. And still, 30% of callers ask for a person. The wingman answers; the human closes the hard ones.

The flip side is brutal. When something goes wrong in an agent-mediated purchase, 86% of those transactions are at risk of being abandoned or switched to a competitor. An agent that gets your check-in time wrong doesn't apologize. It books the hotel next door.

What "legible" actually means

Martin Soler wrote a piece this week about ontology for hotels. Strip the word: it means an inventory of what exists in your property and how it connects. Room 407 is a room type, a floor, a view, a cleaning protocol, a maintenance history, and the thing your housekeeper knows that nobody wrote down.

For a 30-room hotel, that's not a data project. It's a spreadsheet and an afternoon.

• The facts. Room types and what's in them. Check-in and check-out times. What breakfast includes. Parking, pets, cots, the distance to the station in minutes, not "close by".
• The policies. Cancellation, deposits, city tax, minimum stays by season. Written once, in plain language, the same everywhere.
• The 50 questions. The ones your front desk answers every day from memory. If they only live in someone's head, the AI — Google's, the OTA's, or the one answering your WhatsApp — will guess. Or it will use the OTA's version of you.

Two mistakes I see. Over-collecting: nobody needs centimetre-level room data. And under-owning: a single fact — breakfast until 10 or 10:30 — living in three places with three answers. The machine quotes the first version it finds.

Why this beats buying another tool

Only 30% of travel executives say their systems are ready to run a guest journey end-to-end. The chains are spending millions to fix that. You don't have to — you have one property, one team, and every fact within walking distance.

That's the Champions League an independent can actually play in: the chain has more data; you can have cleaner data, faster.

The takeaway: before the next AI tool, write the hotel's answer sheet: the 50 questions guests ask most, one answer each, one owner who keeps it current. Then push that same sheet everywhere a machine reads you — your site, your Google profile, your OTA content, and whatever agent replies to your guests at 11pm. The AI can only answer as well as what you gave it. Give it the truth, once.