AI Travel Risks: Wrong City Bookings Loom Large

When AI Gets Geography Wrong, You End Up in the Wrong Country

LONDON, England - I've spent enough time trying to get out of the wrong city to appreciate a good horror story about travel mishaps. Andrew Jordan, Travelport's chief technology officer, delivered one at last week's TravelTech Show in London that should make anyone excited about AI booking tools pause: there are about 14 places named Birmingham in the United States alone, according to Traveltech Show, and in the shiny new world of AI-driven travel booking, a traveler could conceivably be sent to the wrong one. Not kinda wrong. Not close enough. Actually wrong, as in Birmingham, Alabama instead of Birmingham, UK, or vice versa. The kind of wrong that burns a day of your life and several hundred dollars fixing. Jordan used this example during a session on agentic commerce and travel industry preparedness at the TravelTech Show, which drew over 700 senior travel technology buyers from across the European market, according to Traveltech Show. His point was blunt: the industry is confronting huge change in how products and services are discovered and distributed, and it's moving from deterministic systems that guarantee specific outcomes to probabilistic AI models that generate likely but not certain answers. That shift is not trivial. Probabilistic AI is fantastic for browsing, inspiration, vibes. It's terrible at transactions where precision actually matters.

Why Almost Right Is Completely Wrong

According to PhocusWire, Jordan has been clear that "almost right isn't good enough for travel's agentic future," a statement that reflects the core tension in automating travel booking. AI can serve up plausible options at lightning speed, but plausible is not the same as correct, and in travel, incorrect means real money, real missed connections, real consequences. This is not a theoretical concern. Jordan pointed out that AI-powered agents could eventually fire off up to 100,000 transactions a second through distribution systems, according to PhocusWire. At that scale, even a tiny error rate becomes catastrophic. A one percent failure rate at 100,000 transactions a second is 1,000 travelers sent to the wrong Birmingham, charged the wrong fare, or booked onto flights that don't exist. The issue is fundamental to how large language models work. They predict the next most likely word or concept based on patterns in training data; they don't actually understand intent or verify facts against a ground truth. Ask a probabilistic system to book Birmingham and it might choose based on proximity, popularity, or some weighted average of previous queries. What it won't necessarily do is ask which Birmingham you meant. Traditional booking platforms are deterministic. You input a city code, and the system returns exactly that airport, every time. No guessing, no creativity, no room for misinterpretation. AI doesn't work that way, and that difference is why Jordan and Travelport are pushing for AI to augment rather than replace core booking infrastructure.

What Jordan's Warning Means for Travelers on the Move

I've been using AI tools to brainstorm trip ideas and compare routes for months now, and they're genuinely useful. ChatGPT can suggest offbeat destinations I wouldn't have thought of; generative search surfaces blog posts and forums faster than I could manually dig through them. But would I let any of those tools book a flight without checking every detail myself? Absolutely not. Jordan's Birmingham example crystallizes why. When you're living out of a backpack and working remotely, a mis-booked flight isn't just an inconvenience; it's a blown week of accommodation deposits, missed coworking credits, maybe even a visa issue if you end up in the wrong country. The stakes are higher when you don't have a safety net or a corporate travel department to fix things. According to PhocusWire, Jordan has emphasized that "AI changes discovery in travel; booking is the challenge." That's exactly right. Discovery is low-risk. You browse options, you compare, you discard ideas that don't fit. If the AI suggests something weird, you ignore it and move on. Booking is high-risk. Once that transaction goes through, you're committed, and if the system guessed wrong about your intent, you're stuck untangling it. This is why Travelport is working with partners like Anthropic and Cognizant to build what they're calling an AI travel ecosystem, according to PhocusWire, one that connects AI-driven trip planning to accurate, confirmed bookings travelers can trust. The architecture they're describing keeps AI in the inspiration and search layer but routes actual booking through robust, deterministic platforms with strong data infrastructure. Human oversight remains critical for complex itineraries, disruptions, and edge cases where preferences aren't fully captured in structured data. That's the model that makes sense right now. Let AI handle the heavy lifting of sorting through thousands of options, surfacing personalized suggestions, and automating routine communication. But keep a human or a deterministic system in the final transaction loop, especially for high-value or complex bookings. For budget travelers, digital nomads, and anyone managing their own logistics on the road, the practical takeaway is simple: use AI for research and comparison, but verify every booking detail yourself. Don't assume the system understood your intent. Double-check airport codes, confirm dates, and review the full itinerary before hitting confirm. The tools are getting better, but they're not there yet, and the cost of a mistake is too high to trust them blindly.

The Longer Road Ahead

TravelTech Show hosted more than 100 global suppliers and featured over 548 pre-booked meetings in 2025, according to Traveltech Show, with exhibitors gaining access to decision-makers controlling more than £467 million in collective technology budgets. That level of investment and engagement signals that the industry is taking AI seriously, but it also underscores how much work remains before fully automated, end-to-end AI booking becomes safe and reliable. Jordan's warning isn't anti-AI; it's pro-reality. The technology is powerful, and it will reshape travel in meaningful ways. But travel is not a domain where almost right is acceptable. It's a domain where getting sent to the wrong Birmingham ruins your trip, drains your savings, and destroys trust in the platform that got it wrong. Until AI can match the precision and accountability of deterministic systems, human judgment and strong infrastructure will remain essential safeguards. For now, that means travelers need to stay engaged, stay skeptical, and keep double-checking the details. The future of AI in travel is coming, but it's not here yet, and pretending otherwise is how you end up in Alabama when you meant to be in the Midlands.