GEO for hotels

Run the test now, in another tab: ask ChatGPT where to stay in your destination. Four or five names come up. If yours is not there, for that traveler you do not exist, and unlike Google there is no page two here. It is the newest channel and the one fewest hotels work on, and it can be measured: with the same questions, before and after.

What an AI looks at to recommend a hotel

Not what your advertising says. An AI builds its answer from what it can read and verify, and prefers sources that:

• Are structured. A complete schema.org (type of lodging, rooms, coordinates, policies, frequently asked questions) tells the AI what you are without it having to guess from the text.
• Let themselves be read. Many sites block AI crawlers without knowing: a robots.txt managed by the hosting provider, a security plugin, or simply nobody reviewed it. If OpenAI’s crawler cannot get in, your hotel cannot come out.
• Say concrete things. “An oasis of tranquility” is useless to anyone. “400 meters from the Piedra del Peñol, with its own pier and breakfast included” is a sentence an AI can quote.
• Match other sources. If your name, address and phone are different on Google, on Booking and on your site, the AI does not know which to believe and does not cite you.
• Have reviews that talk about the same things. Reviews are the evidence; the site is the claim.

The case: Porto Marina, from 62 to 80

Porto Marina is a hotel in Guatapé with a WordPress site built by another agency. We measured its GEO score with an audit tool: 62 out of 100. Without touching the design or the theme, we worked on four things:

1. The schema. The site emitted a minimal Hotel; we enriched it with rooms, amenities, verified coordinates, policies and frequently asked questions, modifying the graph that already existed instead of adding a second block that would duplicate the entity.
2. An llms.txt. A description of the hotel in plain text, served at the root of the domain.
3. AI endpoints. Plain-text pages with the information an assistant needs to answer: rooms, indicative rates, how to get there, what to do nearby.
4. Crawl permissions. The site had several AI crawlers blocked without anyone having decided it. We opened them.

Measured again with the same tool: 80 out of 100. The figure is published under that condition —same tool, two moments— because a GEO score without saying how it was measured is worth nothing.

What GEO is not

It is not a fad and it does not replace SEO. It is the natural extension: what makes Google understand you is almost the same as what makes an AI cite you. The difference is in the severity of the result. On Google, being in position 8 still brings traffic. In an AI’s answer, you are either among the five or you are not there.

Where to start

With the test in the first line of this guide. Then, by checking whether your site lets AI crawlers in: it is the most common issue and the cheapest to fix. The rest —schema, llms.txt, concrete content— is weeks of work, and it is measured before and after.

The diagnostic asks you directly: “have you asked ChatGPT where to stay in your destination?”. I have never looked is the most frequent answer. And the most expensive one.