A traveler opens ChatGPT and types: “recommend a hotel in Matera where I can have a truly immersive and authentic experience”. Within seconds, they get three or four names, each with a one-line explanation, already filtered according to their criteria. They choose among those. The others simply do not exist for them. And if your hotel is not among those names, you will never know: no lost impressions to analyse, no missed clicks to recover. Just a guest who never arrived.
The question that matters today is no longer simply “how do I rank on Google?”. It has become: why should artificial intelligence recommend my hotel in particular? The short answer, which I will explore through concrete examples from projects I have personally worked on, is this: AI assistants recommend properties that the web talks about in an authoritative, consistent and clear way. And the most structural way to get the web talking about a hotel has always been PR and press office work.
This isn’t a future scenario: it’s already happening
The numbers no longer leave much room for doubt. According to research by Phocuswright, 56% of active U.S. travelers have used AI to plan, book or receive assistance during at least one trip in the past twelve months, up from 43% in the second half of 2025 and 33% in the first half of the same year. In Italy, the curve is almost identical: according to the Travel Innovation Observatory of Politecnico di Milano, the share of people who used artificial intelligence to organise a trip rose from 15% in 2024 to 33% in 2025, and the vast majority of those who use it consider it useful or even essential.
The problem for independent hotels lies in another figure. The HotelWorld AI Visibility Index, an international analysis of more than 130,000 properties across 30 countries, found that only 16% appear in responses from ChatGPT, Google AI and Perplexity when someone asks where to stay. It is research produced by an industry operator rather than an independent academic study, but the size of the sample still makes the figure significant. The remaining space is taken up almost entirely by the large hotel groups, those with budgets, marketing departments and dedicated teams already working on the channel. In other words: more than eighty properties out of every hundred are invisible precisely where their potential guests are already searching.
Unlike a Google results page, there is no second page to scroll through here. The assistant returns a shortlist of three to five names, presented as recommendations. If you are not on that shortlist, for that traveler it is almost as if you do not exist.
I tested it with my own projects
Before writing anything about this topic, I wanted to see it for myself, starting with the properties I know best because I have worked with them personally. I asked the questions while logged out, so that previous history would not influence the answers.
I asked ChatGPT, Gemini and Perplexity to recommend “a museum hotel or an unusual cultural place to stay in Matera”. All three, without exception, mentioned Moyseion. And not in generic terms: Perplexity described it as a true museum hotel carved into limestone caves, with rooms inspired by the Neolithic period, the Oenotrian civilisation and Magna Graecia, even mentioning its experiential rituals and the historical clothing worn in the common areas. Gemini described it as a “living museum” created according to the principles of experimental archaeology, in collaboration with scholars, artists and archaeologists. ChatGPT echoed the idea of a “museum you can sleep in”.
The same happened with B-Side, the alpine hi-fi listening bar in Alta Badia. When asked about a music-related venue in the Dolomites, all three assistants identified it as the most original option, recognising its connection to Japanese jazz kissa, its bespoke sound system and the “Frequency Menu”.
These are precise, detailed and accurate descriptions. And above all, they are descriptions that an AI assistant can only provide if those pieces of information exist online, repeated consistently across multiple parts of the web.
Where do those answers come from? I asked
This is where things become interesting. I asked the assistants which sources they had used.
Perplexity, transparently, displayed its references directly alongside the answer: the property website, booking platforms and review websites. Gemini initially attributed its answer to its own “prior knowledge” of Matera’s cultural heritage. But when I asked the question directly – whether the many online articles generated through PR activity had been useful – the answer was unequivocal: it explained that the information came from press releases, tourism and cultural trade articles and the media coverage generated around the project. It explicitly acknowledged that PR campaigns and online media presence had amplified its visibility.
There is, however, an important caveat. Asking a model to reconstruct retrospectively where an answer came from does not prove which sources actually determined that output. Gemini’s response should therefore be treated as an interesting indication, not as proof of its internal process.
The tests alone do not demonstrate a direct cause-and-effect relationship, but they do show something interesting: when these assistants reconstruct the identity of these projects, they find online the same narrative elements that PR activity has spread and consolidated over time. In systems that display their sources, part of that process can also be observed directly.
The reason Moyseion appears when a traveler searches for a cultural experience in Matera is not a technical trick: in nine months, the project has been covered by around one hundred articles across twelve countries, in publications including The Guardian, Vanity Fair and Sky Arte. That body of authoritative content has become part of the information ecosystem surrounding the project: a set of independent sources that reinforces associations, distinctive characteristics and context, and that AI-powered search systems can retrieve when they need to formulate a recommendation.
The dam metaphor
When a hotelier asks me how this relationship between media coverage and artificial intelligence works, I answer with the image I use most often.
PR is like opening a dam. From that moment, a river begins to flow: articles generate readers around the world, readers generate searches for the name, searches lead to Google reviews, coverage fuels posts and conversations on social media. It is a flow that, once started, is reflected everywhere – and today it is also reflected inside the answers given by artificial intelligence, because AI drinks from exactly that river.
There is no shortcut for filling a riverbed that is dry. You can work on the banks as much as you like, but if there is no water upstream – if the web has no authoritative reason to talk about you – no amount of technical optimisation will be enough. The dam comes first. Everything else follows.
A word of caution: visible doesn’t mean understood
There is, however, a valuable lesson hidden in those tests, and I would be dishonest not to mention it.
As I continued testing Gemini, at one point the assistant made a mistake: it claimed that Moyseion “is not an accommodation facility or a hotel, but an exhibition and cultural space”, and suggested other places to stay. It was an incorrect description, probably generated because the project’s positioning – “museum hotel”, “living museum” – is so distinctive that it can be misunderstood by a machine.
This tells us something fundamental. Visibility alone is not enough: you also need clarity. If the essential facts about you – what you are, where you are, who you are for – are not repeated unambiguously everywhere, AI can know you and still describe you incorrectly.
That means two things, not one. On the one hand, PR work that does not simply aim to “get coverage”, but consistently reinforces three or four key messages, always the same, until they become the version that the web – and therefore the machine – treats as established. On the other, a clean and coherent website and listings (Google Business Profile, booking platforms), aligned with those same messages, because these are the sources AI can return to when checking basic facts.
The technical side matters, but it’s the floor, not the ceiling
Do not misunderstand me: the technical work matters. I know this from direct experience. After a long and meticulous structural optimisation of the Press & Pillow website – around twenty-four hours of detailed work – I began recording traffic coming from AI assistants. It was worth doing, and I will continue doing it, because this is the direction the world is moving in and those who do not adapt will be left behind.
But it is important to put things in the right order. A website that is accessible to crawlers, consistent information, well-structured content: this is the floor. It makes your hotel readable by the machine. It does not yet make it worthy of being chosen.
It is also worth saying, for the sake of honesty, that some of the “solutions” currently being promoted promise more than they deliver. Google’s own official guidance on generative AI in Search makes clear, for example, that files such as llms.txt are simply ignored by Google Search and that structured data is not the secret to visibility in AI systems: Google specifies that no special structured data or Schema.org markup is required to appear in its generative search experiences. They are useful as hygiene, not as the lever.
The real lever – the one that turns “readable” into “recommendable” – is authority. And authority is built through independent sources talking about you.
Why independent hotels have more to gain
Let us go back to that 16%. Large hotel groups dominate it through budget. But artificial intelligence does not reward spending alone: it rewards authority and distinctiveness. And this is where independent hotels have an advantage that chains cannot simply buy.
A small hotel with a genuine identity and an authentic story can earn media coverage disproportionate to its size – and with it, a level of authority that AI can recognise. Moyseion is not a chain: it is an independent project. It appears in assistant recommendations because it has something real to tell, and because that story has been told well, to the right publications, consistently.
And there is one final piece, perhaps the most important, that no software can replicate. Moyseion’s media coverage has also worked because behind it there is a founder and a team who genuinely believe in the project. When I accompany a journalist, I often do not take them to a restaurant: we have dinner at the owner’s home, informally, so that they can experience not just the place but the people who inhabit it.
That is where the best articles come from – the sincere, detailed ones that are difficult to imitate. A river always begins at an authentic source. If the source is real, the flow can travel a long way, all the way to the answers a machine will give to a traveler on the other side of the world.
What I would do in practice
If you run a hotel and want to start existing in this new channel, this is the order of priorities I would recommend.
Define your key messages and make them unmistakable. Three or four concepts that explain what you are, where you are and who you are for. They should appear consistently on your website, listings and press materials, leaving AI as little room as possible to misunderstand you.
Build authoritative and consistent media coverage. This is the heart of the process: independent sources are what assistants can use to build trust around you. Not one press release, but continuous work that keeps feeding the river over time.
Keep the channels you own in order. A clear and accessible website, consistent information across Google Business Profile and booking platforms. This is the floor: necessary, but not sufficient.
Test your own visibility every month. Open ChatGPT, Gemini and Perplexity and ask for recommendations the way a potential guest would. See whether you appear, which words they use to describe you and which sources they cite. It is the only way to really understand how the machine sees you.
Do not chase tricks. The energy spent on special files or “magic” markup delivers far less than the energy invested where AI actually drinks: authoritative, independent content that talks about you.
In conclusion
Artificial intelligence is not a channel to hack. It is a mirror: it reflects how much, and how well, the web talks about you. You can polish the mirror as much as you like, but what it shows depends on what you have built in front of it.
And who decides what the web says about a hotel – which stories, which publications, which level of consistency – is, to a large extent, the work of PR. It is no coincidence that the projects receiving the most meaningful coverage are also among those that machines are more likely to recommend. Visibility in AI is not the starting point: it is the reflection of authority built elsewhere.
If you believe your hotel has an authentic story but still does not appear where your guests are searching, the first step is not a technical intervention. It is understanding what story you can offer the media and how to turn it into that continuous flow which, over time, can travel everywhere – artificial intelligence included.
If you want to understand its editorial potential and how to build an authoritative media presence – and, over time, visibility within the new AI-powered search environment – we can discuss it in an introductory call.
