All articles

Journal 6 min read 9 November 2026

Google and AI can only use what your website states.

Before you pay anyone to make your hotel visible to AI, run a ten-minute test. Ask the questions your guests ask, then check whether your own website answers them in plain text. Here is the test, what Google actually requires, and what a website can't fix.

A guest asks an AI assistant for a quiet hotel near the lake, with parking and breakfast until eleven. The answer names three places. Whether yours is one of them depends on many things, and most of them are not your website. One of them is. Before you pay anyone to make your hotel “visible to AI”, it’s worth ten minutes to see what your own site actually tells a machine.

What changed, in one sentence

Search now answers questions itself. In March 2025, around one Google search in five already showed an AI summary, and people who saw one clicked through to a website in 8% of visits, against 15% when there was none (Pew Research Center). Fewer visits means the answer matters more than the click. And the answer can only use facts it can find.

What Google actually asks for

There is a small industry selling “AI optimisation” for hotels. Google’s own guidance is plainer. For AI Overviews and AI Mode, it says there are “no additional requirements” and no “special optimizations necessary”. It also says you “don’t need to create new machine readable files, AI text files, or markup”, and that “there’s also no special schema.org structured data that you need to add” (Google Search Central).

What it does list is ordinary: let Google crawl the site, put important content in text, make sure structured data matches what’s visible on the page, and keep your Business Profile up to date. None of that is new. Most hotel websites still miss some of it.

A website can’t make an AI choose you. It can make sure the AI has nothing wrong to say.

The plain test

One. Write down five questions guests actually ask before booking: check-in time, parking, dogs, breakfast hours, how to get there without a car. Ask them about your hotel in ChatGPT and in Google. Note every wrong or missing answer, and where it seems to come from.

Two. Open your own website and look for each answer as text on a page. Not inside a photo, a PDF menu, or the booking engine’s window, but in words a person could copy. If you can’t find it in under a minute, neither can a machine.

Three. Paste your home page into Google’s Rich Results Test or the Schema Markup Validator and check your hotel schema markup. You’re checking two things: that structured data exists, and that it says the same as the page. A check-in time of 14:00 in the code and 15:00 on the page is worse than no code at all.

Four. Check that your robots.txt doesn’t block Google, and that your pages exist in every language you sell in, with the language marked. A German guest asking in German should find German text.

Five. Compare your website with your Google Business Profile and your main OTA listings. Same phone number, same check-in time, same breakfast hours. Contradictions are where wrong answers start.

What a website can’t fix on its own

Here is the honest part. A study by Stiplo of 659 Barcelona hotels, hostals and guesthouses, measured across more than 27,000 AI travel answers and presented at Skift’s Data + AI Summit Europe in October 2026, found that independent hotels were far more likely than chain hotels to be missing from AI answers altogether. The strongest signal it found was Google review volume. Technical extras such as schema, llms.txt and AI-crawler settings showed no independent effect in that data (Stiplo; Skift).

Two things from the same study matter for your website. A hotel’s own site was the source of only a small share of AI recommendations. But hotels with pages answering real guest questions were visible 87% of the time, against 72% without them. Practical pages helped. Clever code didn’t.

So a website is one part of the answer, not all of it. Reviews, your Business Profile and your OTA content carry a lot of weight. If you want all of them checked together, that is a full Search & AI Visibility Audit, which Marco runs through Rosso Consulting.

What we do on every Astia site

We state the facts guests ask about as plain text on the page, in every language the site has. We add structured data that matches that text, a sitemap and language tags, and we keep the site open to Google and the main AI assistants, and to the booking engine, channel manager and CRM. Pages are static and fast, and must pass Lighthouse 90 before launch. We also publish an llms.txt file, because it costs nothing, but we don’t promise it does anything. What helps and what doesn’t, with the evidence for each point, is on our page GEO, explained honestly. You can see what a guest sees and what a machine reads side by side on how it works. What that costs over five years, including the case where a lean one-off is cheaper, is on our pricing page.

When the rules change, we change the site. That’s part of the 150, like everything in our five-year cost comparison. What we won’t do is promise an AI ranking. Nobody can.

The decision for Monday

Run the five questions. If your website can’t answer them in plain text, fix that first, whoever builds your site. If you want a site that states those facts in every language and keeps the structured data matched to the page, book a 30-minute call. If the site already can and the AI still gets you wrong, the problem is probably somewhere else, and that is the full Search & AI Visibility Audit through Rosso Consulting.

Want to see our articles first when you search? Add Astia Web as a preferred source on Google

Marco Rosso, founder of Astia Web, in a navy blazer and striped shirt.

Marco Rosso

Founder, and your contact at Astia Web

Ten years in hospitality technology.

Want this applied to your own site?

Thirty minutes, no pitch. We look at your site and tell you plainly what we would change, and what we would leave alone.