— Mirko Lalli
The Iran conflict has now canceled over 23,000 flights and effectively shut down Dubai, Doha, and Abu Dhabi as global transit hubs. Fortune calls it the largest disruption to the $11.7 trillion travel industry since COVID. That comparison undersells the difference: COVID was a demand shock we could see coming; this is a supply shock with no clear end date, and jet fuel prices have doubled since February.
My read is that we are watching a permanent rerouting of global travel flows, not a temporary pause. British Airways and Cathay Pacific are already adding capacity on alternative Asia-Europe routes, which tells you the majors expect this to last. United's Scott Kirby is signaling ticket price increases tied to fuel costs, and The Points Guy's team is advising travelers to abandon traditional booking windows entirely. When industry voices say "book now, don't wait," they are pricing in further deterioration.
For DMOs and hoteliers outside the conflict zone, the next six months will bring redirected demand but also redirected cost pressure. European and Southeast Asian destinations should expect both opportunity and sticker shock as travelers reroute but pay more to get anywhere. The cost-of-living squeeze means leisure travel budgets are shrinking just as getting from A to B gets more expensive. Tech vendors selling revenue optimization tools will find a receptive audience, but only if they can model volatility rather than assume stable baselines.
The strategic question for any travel business right now: are your forecasting models built for a world where major hubs can go offline in a week?
Forty-two percent of travelers used ChatGPT for trip planning in 2025. That number comes from COAX Software's analysis of travel technology trends. It means that in less than two years, a tool that didn't exist became the default starting point for nearly half of travel decisions.
The travel industry has responded with a building spree. Agentic AI systems. Autonomous rebooking. Predictive interventions. Zero-touch travel. The narrative from every platform vendor is the same: we're removing friction. We're making it seamless. We're letting the machines handle it.
But here's the question nobody at the trade shows is asking: What if friction was the product?
The Conversion Obsession
According to Appinventiv, AI-powered flight booking systems are now achieving 35% higher conversion rates through fare prediction, smart search ranking, and route optimization. This means that for every hundred people who would have abandoned a booking, thirty-five more are now completing it.
The industry celebrates this as progress. More bookings. Higher efficiency. Better outcomes.
I'm not so sure.
Let me be clear about what these systems actually do. They analyze your behavior. They rank results not by what's objectively best, but by what you're most likely to click. They time their offers to hit you when the data suggests you're most vulnerable to a decision. Alaska Airlines' tool, cited in the same report, personalizes results based on your browsing history, your past purchases, your apparent price sensitivity.
This isn't neutral infrastructure. It's persuasion architecture. And the 35% conversion lift measures how well it works.
Now, I spent years building data systems for the travel industry. I understand the value of relevance. Showing someone flights they can actually afford, at times that work for their schedule, from airports they can reach — that's useful. That's service.
But there's a line between relevance and manipulation. The current AI race has blurred it beyond recognition.
The Zero-Touch Promise
According to a joint report from Deloitte and Data Appeal (my former company, disclosure noted), Agentic AI is the defining trend in 2026 travel tech. These systems can make autonomous decisions — rebooking flights, reconciling expenses, adjusting itineraries — without human input.
The industry calls this zero-touch travel. I call it something else: zero-agency travel.
The pitch is seductive. Your flight gets cancelled, the AI rebooks you before you even know there's a problem. Your hotel rate drops after booking, the AI claims the refund automatically. Your connection is tight, the AI pre-clears you for the faster security line.
All of this sounds wonderful until you realize what it requires: handing over complete control of your travel experience to systems optimized for someone else's objectives.
Here's the camera flip that the vendor presentations never show. Imagine you're not the traveler. You're the AI.
Your training data comes from millions of completed journeys. Your reward signal is whatever the platform defined as success — bookings completed, revenue captured, complaints minimized. You have no concept of why someone is traveling. You don't know that the business trip to Chicago is actually a chance to see a dying parent. You don't understand that the "inefficient" routing through Lyon has a reason — a friend who lives there, a restaurant that matters, a memory being chased.
You optimize for the measurable. The meaningful never makes it into your training data.
The Juniper Paradox
Juniper Travel Technology's JuliA platform exemplifies the current state of the art. According to their documentation, it provides real-time data analysis, personalized recommendations, and streamlined customer service. The system learns from every interaction, continuously improving its ability to predict what travelers want.
But what JuliA actually learns is what travelers accept.
These two things are not the same. What I want and what I accept diverge constantly. I want a direct flight. I accept a connection because it's cheaper. I want a hotel with character. I accept a chain because I'm exhausted and need predictability. I want to discover something unexpected. I accept the algorithm's recommendation because discovering takes effort.
Every AI system trained on acceptance data is learning to optimize for surrender, not satisfaction.
The vendors will object. They'll point to satisfaction scores, to repeat bookings, to reduced complaint rates. But these metrics measure whether travelers are annoyed enough to act, not whether they're genuinely served.
I ran customer satisfaction programs for years. Here's what I learned: the absence of complaint is not the presence of delight. Frictionless is not the same as wonderful.
The Saudi Experiment
Saudi Arabia is making a $135 billion bet on AI-powered tourism, according to Artificial Intelligence News, projecting that contribution to GDP by 2030. They're using AI for cultural tourism promotion and sustainability optimization at scale.
This is the most interesting case study in the current landscape, precisely because it's the most transparent about its objectives.
Saudi Arabia is not pretending that AI is neutral infrastructure serving traveler preferences. They're explicitly using it as a promotional tool — to shape perception, to direct attention, to construct a narrative about cultural experience.
At the TOURISE Summit, they're also discussing ethical frameworks for travel tech AI. This combination of aggressive deployment and ethical handwringing perfectly captures where the industry is right now.
Everyone knows the current trajectory raises uncomfortable questions. Nobody wants to slow down enough to answer them.
The MEXC Reality Check
According to MEXC's analysis of travel technology platforms, AI enables faster bookings by checking real-time availability across suppliers, smarter dynamic pricing based on demand patterns, and hyper-personalized recommendations tailored to user habits.
Let's decode that.
Real-time availability checking across suppliers is genuinely useful. It's plumbing. It makes the system work better without trying to influence your decisions.
Dynamic pricing based on demand patterns means prices change based on how badly the system thinks you need something. The AI's job is to find the maximum you'll pay, then charge you slightly less than that.
Hyper-personalized recommendations tailored to user habits means the system creates a filter bubble around you, showing you more of what you've already consumed and less of what might challenge or surprise you.
Two of these three things are not in your interest. They're being sold to you as features.
What Nobody's Building
Here's my contrarian position: the most valuable AI application in travel hasn't been built yet, because nobody's willing to pay for it.
An AI that tells you not to book. That recognizes when you're stress-shopping for a vacation you don't have time to take. That surfaces hidden costs before you commit. That shows you what you'll actually experience versus what the photos promise. That recommends the cheaper option when the expensive one won't make you happier.
This AI would destroy conversion rates. It would reduce transaction volume. It would cut revenue per session.
It would also build trust that lasts decades.
The reason nobody's building it is structural, not technical. Platform economics reward transactions. Every investor deck, every board meeting, every quarterly review asks the same question: how do we increase bookings?
Not: how do we ensure the bookings that happen are the right ones?
The 35% conversion lift is celebrated as progress. But some percentage of those converted bookings are trips people shouldn't have taken. Vacations that won't deliver. Business travel that could have been a video call. Experiences that sounded better in the algorithm than they felt in the flesh.
The AI captured the booking. The human paid the cost.
Where This Leaves You
If you run a DMO, a hotel, or a travel business, you have a choice to make.
You can adopt these systems and celebrate the conversion lifts. You can call it innovation and collect your efficiency gains. You can optimize for the measurable and assume the meaningful will take care of itself.
Or you can ask harder questions.
What do your travelers actually want, not just what do they accept? Where is friction serving them, not just slowing you down? What would it mean to optimize for satisfaction that shows up years later, in return visits and word-of-mouth and trust?
The AI vendors will keep building faster autopilots. The question is whether you want your travelers to arrive somewhere they chose, or somewhere the algorithm decided they would accept.
I use AI as a thinking partner and to polish my English. The analysis and positions are mine.