Insights

AI Discovery · Direct Bookings

Is your hotel invisible to ChatGPT?

Only about 16% of the world's hotel supply appears in answers from ChatGPT, Gemini and Perplexity. How to check whether yours does, in ten minutes, and what actually changes the outcome.

Directel

Only about 16% of the world's hotel supply appears in the answers given by ChatGPT, Google's AI and Perplexity. The figure comes from the World's Best at AI 2025 Index published by HotelWorld AI, and it is worth sitting with for a moment. For roughly five properties out of six, a traveller who asks an assistant to plan a trip will simply never hear the name.

This is not a forecast about the future of search. It describes a channel that already exists and that most hotels cannot see, because nothing in their reporting shows it.

Guests stopped searching. They started asking.

For twenty years the shape of hotel discovery was stable: a query, a list of results, a comparison, a booking. Every tool the industry built assumed that shape.

Conversational assistants break it. A traveller no longer types "boutique hotel Florence". They ask for a quiet place near the centre, walkable to the Uffizi, good for a couple arriving on a late train, under a certain budget. The assistant returns three or four names with reasons attached. There is no page two, no list of twenty options to scroll through, no chance to appear further down.

The consequence is uncomfortable and simple: in this channel visibility is not a spectrum. You are named, or you are not.

Why most properties are absent

Three reasons come up again and again, and none of them are about advertising budget.

Inconsistent data. Assistants cross-check what they find. When the address on your website differs from the one on your Google profile, when the room count varies between platforms, when the name appears in three spellings, the model has no confident answer to give and moves to a property it can verify. Consistency is not a detail here. It is the entry ticket.

Thin presence on sources the model trusts. Assistants lean on places where information is structured and corroborated: mapping services, established directories, well-known review platforms, editorial coverage. A property that exists almost exclusively inside booking portals is visible to those portals and nearly invisible outside them.

Sites written for people, not for machines. Beautiful photography and an evocative paragraph do not tell a model that there is a lift, that check-in starts at three, that the property is a fifteen-minute walk from the station. If those facts live only inside an image or a PDF, they do not exist.

How to check yours, in ten minutes

Do not take anyone's word for this, including ours. Open ChatGPT, Gemini and Perplexity and ask each of them the way a guest would.

Start broad: "Recommend three boutique hotels in the centre of Florence for a couple, within walking distance of the Duomo." Then narrow it to whatever makes your property distinctive: "Which hotels in Florence have a rooftop terrace and allow small dogs?" Then ask directly: "Tell me about [your property name]."

That last question is the most revealing. If the assistant knows nothing, you have a visibility problem. If it answers with details that are wrong or out of date, you have a worse one, because incorrect information is being handed to travellers with the confidence of an answer.

Write down what comes back. That page of notes is the most useful competitive research you will do this quarter, and it costs nothing.

What actually moves the needle

The work is unglamorous and consists mostly of removing ambiguity.

Align your core facts everywhere they appear: name, address, contact details, room count, category. Make sure the practical information guests actually ask about — check-in and check-out times, parking, accessibility, pets, walking distances to the places people go — is written as plain text on your own site, not buried in a downloadable file.

Add structured data to your website so the same facts are machine-readable. This is the one genuinely technical item on the list, and it is a one-off job.

Keep collecting reviews, and keep answering them. Assistants weigh corroboration, and a property with recent, consistent feedback across several platforms is easier to recommend than one with a handful of old entries.

Finally, publish something with substance. Not marketing copy: the answers to questions your guests actually ask. A page that explains the parking situation on your street honestly is far more likely to be surfaced than a paragraph about your passion for hospitality.

What not to expect

Nobody can guarantee placement in an AI answer, and anyone selling you that guarantee is selling something else. There is no auction to enter, no equivalent of a paid position, and the systems change without notice.

What you can do is remove every reason a model has to skip you. That is a different promise, and it is the honest one.

There is also a timing argument that will not last. Most hotels have not started, so the effort required to stand out is unusually low right now. That window closes as soon as the industry notices, which historically takes about eighteen months.

The uncomfortable part

If a traveller asks an assistant for a hotel in your city tonight and your name never comes up, nothing appears in your analytics. No lost session, no bounce, no line in any report. The demand simply goes elsewhere, silently.

That is what makes this different from every other visibility problem the industry has faced. You cannot manage what you cannot see, and the first step is not a strategy. It is opening the assistant and asking.


Published July 2026. Guidance on AI discovery changes quickly; we review this article every six months.

— Ready to grow direct?

Book a free consultation with the Directel team.

Book a consultation