My first Phocuswright was in 2013, in Los Angeles, and the theme of the day was travel personalisation through technology. Everyone spoke brilliantly about it, with impeccable slides and professional optimism. Then not a lot actually happened. For fifteen years.

At DestinationLab 2026, in Rimini on 28 April, the keynote was called "DMOs: Change or Disappear in the Age of Artificial Intelligence." It started from a tweet. The one from 30 November 2022, in which Sam Altman wrote "today we launched ChatGPT." 1.1 thousand comments, 7K retweets, 29K likes. That tweet is the moment that will be remembered as the French Revolution of our time. Not because it's a flashy metaphor. Because from that day the world stopped being as it was before, and none of us have really managed to keep up with it.

What follows is what I said that morning, and what came out of the panel that followed.

The superpower

To explain the superpower of AI I use the story of the Irish guy who paid eight euros for a pint of Guinness in a Dublin pub and, quite reasonably, got a little annoyed. In a couple of days, by himself, he built a website. It's called Guinndex. A system of AI agents that phones every Irish pub, asks the price of the beer and updates it in real time. 6,155 pubs, 1,559 prices verified and geolocated by county. The average price of Guinness in Ireland has fallen since the site launched. Transparency does that.

Five years ago, that project would have been a startup. It would have needed investors, a team, a business model, probably a year to get off the ground. He did it in a few days.

That is the superpower of AI. Eliminating the friction between the idea and its execution. Obviously you still need to have the ideas.

The problem that already existed

67% of travellers say they experienced some form of frustration during the trip planning and booking phase. Two thirds. Organising a trip is complicated. It was.

AI didn't create the change. It solved a real need that already existed.

Old research from Expedia found that in the five weeks before booking a single hotel, the average user consulted 285 different sources. 285. Real work. And it's a slippery slope: when something simpler comes along, people choose the simpler thing. It's always been that way.

Today that journey has shortened considerably. Not ten results to scroll through, not twenty hotels to compare. You ask, and the AI answers with two recommendations. Two. This changes pretty much everything, starting with marketing.

From searching to asking

What I call the silent revolution: for technologies to be truly useful they have to disappear, and AI does it better than any technology before it. We've been using it for sixty years without knowing it — inside Netflix, inside Google, very likely inside your washing machine. The more invisible it becomes, the more we use it without thinking.

The first shift is from searching to asking. Phocuswright published the data: 59% of people already use AI to plan a trip. And 41% of Gen Z and Millennials trust the recommendations of an AI agent more than those of friends or social media. That's from Kayak. Pretty staggering.

Even trust — which has always been the real currency of tourism — is repositioning itself.

There is, however, some good news: the AI channel has the highest conversion rate of any marketing tool. 50% of AI searches become bookings. For traditional web marketing you're looking at a few percentage points, if you're good. Double any other channel — but you have to be there.

Wikipedia has lost 23% of its traffic in the last 18 months. One billion, one hundred and sixty million fewer page views every month. People search on Google, see the little summary at the top of the page, and don't click through anymore. In the same period, though, AI traffic to travel sites in the United States has grown by 3,500%. The destination websites I know are recording a drop in organic traffic. They're declining. Not because people have stopped searching — because they no longer click.

From asking to delegating

The second shift has already started. It won't be long before, beyond asking AI for advice, we'll say "okay, you handle it," "book everything." After all, you already know what I like, how I travel, the brands I prefer, and so on.

The new asymmetry

It's no longer a question of big versus small, enormous budgets versus minimal ones. The new asymmetry is between those who speak the language of machines and those who don't. Between those who are machine-readable and those who are machine-invisible.

NerdWallet is an American personal finance website: credit cards, mortgages, insurance. It has a section on credit card travel benefits — airport lounges, hotel discounts. A minor subsection, born for other purposes. An American firm ran an analysis: when you search on ChatGPT for something related to Hyatt hotels (mentioning the brand), NerdWallet appears in 13% of cases. Hyatt's own official website appears in 9%. A finance site is cited more often by AI models than the official site of one of the world's most important hotel chains.

Shifting from an SEO mindset to a GEO or AEO (Generative Engine Optimisation / Answer Engine Optimisation) mindset isn't an overly complicated technical question. Yes, there are some technical interventions involved, but it's mostly about how you write content, how you structure data, how you give access to AI bots. Some Italian destinations block everything. They block exactly those bots that could find them and recommend them at the moment a traveller is deciding where to go.

The destination of the future should operate on three levels. The first is technological: structured data, semantics, knowledge graph. The invisible foundation. If machines don't understand you, they don't recommend you. The second is the traveller level: conversation, vibe, personalisation. Not "Tuscany November" but "quiet hilltowns in Tuscany in November with food fairs." AI has normalised what tourism claimed to want to do for twenty years — having a truly personalised offer. The third is the community level: identity, authenticity, values, balance. The one thing an AI cannot invent and cannot replace.

The panel after the keynote

"Overcoming seasonality in tourism: the integrated role of data, technology and hospitality culture in light of the AI revolution." Three voices.

Giulia Dalla Palma, Head of ATA & Destination Development at Trentino Marketing, talked about the "beautiful seasons" strategy: spreading flows outside peak periods to generate value in the valleys and local communities, driven by both economic and territorial imperatives. The most interesting part: in Trentino, the tourism law has included coaching for operators as a system-wide objective. Not a project — a mandate written into the legislation. They're working on an AI assistant for the internal platform, on semantic reading of content, on a generative Tonal Voice. I don't have many examples of Italian destinations moving with this kind of systematicity.

Andrea Monti, CEO of EHL Next, the Innovation Hub of EHL in Lausanne: EHL is the world's first school of hospitality management, founded in 1893 by a consortium of hoteliers who wanted to train staff. Students in the first six months run a starred restaurant, do room service, manage the front desk. Empathy with the guest is learned that way, not in a lecture hall. The message was clear: practical implementation matters more than rapid adoption. One size doesn't fit all. Large chains can afford all-encompassing technologies. Operators in smaller destinations need to do more artisanal work — slower, more specific. For us Italians, he said, this could be an advantage in the long run.

Davide Filiaggi, co-founder of RankWit.ai: the idea for the startup came to him when he noticed organic traffic on a website falling while traffic attributed to ChatGPT referrals was growing — with a much higher conversion rate. That question led to a product. The SEO approach is deterministic: you're first, second, third, and you stay there. The GEO approach is probabilistic. There's a probability that a model will mention you or won't, and that probability shifts depending on content quality, schema markup, and bot access. RankWit Hotel for hoteliers who want to understand how they're cited in AI models and reduce their OTA dependency. RankWit Destination for DMOs and consortia. The number he closed with: 80% of travel results on ChatGPT come from OTAs. The OTAs have invested billions to dominate this channel. Hotels, destinations, consortia haven't done it yet. The window is open, but it won't stay open forever.

Humanity as superpower

In a world where AI manages logistics, data, bookings and customer service, interacting with a real person becomes something special. For Phocuswright, 75% of travellers still prefer a human being for guided tours. 76% prefer a human being to take their order at a restaurant.

And once you've solved the visibility problem with AI agents — automate everything that can be automated. Have AI assistants answer the phone, manage bookings, write the FAQs. Free up your people from logistics. Then put those people back where they matter: in the relationship, in the empathy, in the care of guests. That's the one thing the models don't yet know how to do.

The superpower isn't AI. It's you.

***

My DestinationLab presentation is available for download here: https://mirkolalli.me/dlab2026