In a clip now making the rounds on LinkedIn, Airbnb's CEO made what he called a "radical statement" about the future of AI in travel. He's half right.

The Clip That Got Travel Tech Talking

In a recent interview making the rounds on LinkedIn, Airbnb CEO Brian Chesky made what he called a "radical statement": "I do not think a chatbot is the right interface for travel or e-commerce."

He listed four problems with chatbots: they're text-based, they lack direct manipulation, they make comparison shopping hard, and they're "single player," not collaborative.

Then he said something we actually agree with: "The future are agents. But I don't think they're going to be text-forward. I think they're going to be really rich user interface."

Here's where Chesky is asking the wrong question. He's framing this as a front-end problem, focused on what the traveller sees on screen. But the real bottleneck in AI-powered travel booking isn't the interface. It's the supply side. And the data proves it.

The Interface Debate Is a Distraction

Chesky's critique of chatbots sounds reasonable until you look at what travellers are actually doing. Nearly half have already used AI chatbots for travel assistance, and a majority prefer AI-enabled booking platforms over traditional ones (WiFi Talents). The broader numbers are even more telling:

• 40% of travellers globally already use AI-based tools for trip planning and booking (Travala, 2025)

• 80% of travellers are open to using AI for planning and booking (Statista)

• AI-driven personalization increases booking conversion rates by 18-25% (WiFi Talents)

These aren't projections. They're current adoption numbers. Travellers aren't waiting for Chesky's "completely different interface." They're already booking through conversational AI, and they're converting at higher rates when they do.

Chesky's Four Problems: Solved by the Supply Side, Not the UI

"It's text-based. Photos are an afterthought."

This is true of a generic ChatGPT prompt. But it's not an inherent limitation of conversational AI. When a booking platform stores structured inventory (photos, pricing tiers, real-time availability, operator details), any AI system can pull and present that data however it wants. Rich cards. Image galleries. Interactive calendars. The constraint isn't the chat window. It's whether the underlying data is structured and accessible in the first place. Most tour and activity operators don't have that. Their "inventory" is a PDF price list and a WhatsApp number.

"No direct manipulation - you have to type every prompt."

This conflates the current state of chat UIs with their ceiling. Every major AI platform now ships rich interactive elements beyond text: buttons, toggles, embedded apps, visual canvases. The "just text" era ended in 2024. But more importantly, for tours and activities, most bookings are relatively simple. Pick a date, pick a time, pick the number of people. That's three inputs. You don't need a custom-built app for that.

"Hard to compare - you usually want to see more choices."

This is the strongest point, and it's exactly where supply-side infrastructure matters most. Comparison shopping requires structured, normalized data across multiple providers. Today, tour and activity inventory is fragmented across hundreds of operator websites, each with different formats, availability systems, and booking flows. No AI can comparison-shop what it can't access. The solution isn't a fancier UI. It's a standardized data layer that makes every operator's inventory queryable by any AI system.

"Most AI is single player - not collaborative."

Fair point, but this is a feature gap in current AI products, not a fundamental limitation of conversational interfaces. Shared chats, collaborative threads, and multi-user sessions are all technically trivial. Slack solved this in 2013. The real question is: when that shared planning session concludes with "book it," can the AI actually complete the transaction? That depends entirely on whether the supply side is connected.

The Real Problem: 90% of Tour Inventory Is Invisible to AI

Here's what Chesky doesn't address, and it's the real issue for anyone outside the hotel and flights space.

The tours and activities sector is a $270B+ market dominated by small, independent operators. Most of them run their booking through a combination of phone calls, WhatsApp messages, and basic websites with PDF price lists. Their inventory is invisible to AI agents, not because the AI can't handle it, but because there's no structured data to access.

This is the actual bottleneck. You can build the most beautiful, immersive booking interface in the world. If a kayak tour operator on the Algarve coast still manages reservations through a notebook and a phone call, no AI, conversational or otherwise, can book it. The data simply doesn't exist in a format any software can read, let alone an AI agent.

This is the problem a new wave of booking platforms is solving. When an operator moves from WhatsApp and spreadsheets to a system that structures their availability, pricing, capacity, and customer data, their inventory becomes visible and bookable: by their own website, by OTAs, and increasingly, by AI systems that can query structured data. The more operators that digitize, the more inventory becomes accessible to whatever interface the future brings.

That's not a chatbot. That's infrastructure. And it works regardless of whether the front-end is a chat window, a rich visual app, or whatever "completely different interface" Chesky has in mind.

The Industry Is Betting on Conversational, With Data to Back It

The broader industry data backs this up. A majority of travel leaders already see chatbots and virtual assistants as the most significant AI application in travel (AIMultiple), and chatbots handle up to 70% of customer service interactions at major travel companies (Gitnux). But the investment numbers tell an even bigger story:

• The AI-in-tourism market was valued at $3.37B in 2024 and is projected to reach $13.87B by 2030, a 26.7% CAGR (Grand View Research)

• Nearly 30% of travelers now use AI for full trip planning, not just research (Criteo, 2026)

• 83% of travellers are more likely to book when AI-enhanced services are offered (WiFi Talents)

PwC's 2025 report on agentic commerce in travel makes it explicit: the future is not about the interface. It's about AI agents that can take action. Plan, book, pay, modify, cancel. The doing is what matters, not the window dressing.

What Chesky Gets Right (and Why It Actually Supports Our Thesis)

To be fair, Chesky nails one thing: "The future are agents." We completely agree. Agents, autonomous AI systems that act on behalf of users, are the next paradigm for travel booking.

But agents need rails. They need structured inventory to query, real-time availability to check, and transaction APIs to book through. Without that supply-side infrastructure, agents are just recommending things they can't deliver.

Chesky is building Airbnb's agent layer on top of Airbnb's own walled-garden inventory. That works for accommodations where Airbnb controls the supply. But for tours, activities, and experiences (the fastest-growing segment of travel), supply is radically fragmented. No single platform controls it.

That's why the winning approach isn't building a prettier interface. It's building the booking rails that connect fragmented supply to every AI agent simultaneously. More operators on the rail = more inventory for AI agents = more bookings = network effects.

The Bottom Line

Brian Chesky is right that today's chatbots have limitations. But he's answering the wrong question.

The question isn't "What should the AI booking interface look like?"

The question is "Can the AI actually book the thing?"

For hotels and flights, which are largely digitized and API-connected, the answer is increasingly yes. For tours, activities, and experiences, it's mostly no. Not because of the interface, but because of the infrastructure.

But that's changing fast. As more operators digitize their inventory, a tipping point is coming: the moment when AI agents can access enough structured supply to make real-time, end-to-end booking the default for tours and activities, not just hotels. When that happens, it won't matter whether the interface is a chatbot, a visual app, or something entirely new. What will matter is which operators are connected to the rail, and which are still invisible.

CaptainBook helps tour and activity operators digitize their bookings and grow their business. Operators across Europe and the US trust us to manage their availability, pricing, and customer experience. Book a demo to see how it works.

Sources

• Travala - AI travel booking adoption (2025)

• WiFi Talents - AI in global travel industry statistics

• Statista - Travellers open to using AI (2025)

• Grand View Research - AI in tourism market report

• AIMultiple - Travel chatbot research

• Gitnux - AI in travel industry statistics

• Criteo - Summer travel trends 2026

• PwC - The future of agentic commerce for travel (2025)

• PhocusWire / Mary Meeker - AI trends report 2025