— Mirko Lalli
Airbnb now generates 60% of its new code with AI. That number stopped me cold, not because it's surprising but because it makes the workforce debate feel suddenly concrete.
This week's signals cluster around a single theme: the agentic layer is arriving faster than most travel companies are preparing for it. Google Cloud used cruise booking to demonstrate what agentic AI actually does in practice. Mindtrip launched Sabre's in-chat flight checkout. GetYourGuide is targeting the gap between planning and booking with AI that doesn't wait for you to search. Meanwhile, Skift reports that 94% of hotels remain invisible to AI search results, which means the distribution game is being rewritten while most players haven't even entered. Microsoft's new guide on AEO and GEO makes the stakes plain: the $90 billion SEO industry built for human eyeballs faces real disruption as consumers shift to AI-powered discovery. Luxury brands, per Skift, are waking up to the fact that their next booking may come from an AI agent, not a human browsing their website.
For DMOs and hoteliers, the next twelve months demand uncomfortable choices. You need content structured for machine readability, not just human persuasion. You need your inventory accessible to agentic systems that book without human confirmation. And you need staff who understand this shift, which is why the upskilling gap Skift identified worries me more than automation itself.
The question I keep returning to: if your property doesn't exist to an AI agent, does it exist to the next generation of travelers?
Ninety-four percent of hotels are invisible to AI search. Not poorly ranked. Not underperforming. Invisible. According to Skift, when travelers ask ChatGPT or Perplexity for hotel recommendations, the vast majority of properties simply do not exist in the response. The AI has never heard of them.
This is not a marketing problem. This is an existential one.
The travel industry spent two decades optimizing for Google's blue links. We learned SEO, mastered meta tags, paid for AdWords, and built entire revenue management strategies around OTA rankings. Now the rules are being rewritten mid-game, and most of the industry is still playing by the old ones.
This week's news makes one thing clear: AI is no longer a feature travel companies add to their stack. It is becoming the stack itself. And the speed of this transition is catching even sophisticated operators off guard.
The Sixty Percent Threshold
Airbnb announced that artificial intelligence now generates sixty percent of its new code. According to The Tech Buzz, this shift allows the company to accelerate development cycles and respond faster to market demands.
Pause on that number. Sixty percent.
This is not AI assisting developers. This is AI doing the majority of the building while developers supervise, correct, and direct. The human role has shifted from creator to editor, from builder to architect.
The implications for travel tech are immediate. Airbnb can now ship features faster, test more variations, and iterate at a pace that traditional development teams cannot match. For competitors still running conventional engineering organizations, this creates a compounding disadvantage. Every quarter Airbnb operates at this speed, the gap widens.
But here is what most commentary misses: the sixty percent figure is a snapshot of a moving target. If current trajectories hold, that number will be seventy percent in eighteen months and eighty percent in three years. The question is not whether AI-generated code will dominate software development. The question is what happens to companies that reach this threshold last.
The Booking Interface Disappears
Mindtrip launched Sabre's Agentic Flight Booking this week, featuring in-chat checkout capabilities. According to Skift, travelers can now complete an entire flight booking—search, selection, payment—without ever leaving a conversational interface.
This sounds incremental. It is not.
For thirty years, the travel booking interface has been fundamentally visual: grids, calendars, dropdown menus, comparison tables. The mental model required travelers to translate their intent into structured queries. "I want to visit somewhere warm in February" became a series of destination searches, date comparisons, and price filters.
Agentic booking inverts this. The traveler states intent in natural language. The AI translates that into structured queries, retrieves options, presents them conversationally, and handles payment. The traveler never sees the underlying complexity.
This is where the camera needs to flip.
From the traveler's perspective, this is liberation. No more learning each OTA's interface quirks. No more opening twelve browser tabs to compare prices. No more decoding the difference between "flexible dates" and "lowest fare calendar." You say what you want. You get options. You book.
From the hotelier's perspective, this is terrifying. When the booking interface disappears, so does your ability to influence the decision through design, positioning, or promotional placement. Your property becomes a data point in someone else's recommendation algorithm. Your brand becomes whatever the AI decides to say about you.
The hoteliers who survive this transition will be those who understand a hard truth: in an agentic world, your website is not your storefront. Your structured data is your storefront. Your API is your sales team. Your machine-readable content is your marketing department.
The Invisible Majority
Return to that ninety-four percent figure. According to Skift's analysis of AI search results, the overwhelming majority of hotels do not appear when travelers use AI tools for trip planning.
The obvious response is to ask how properties can become visible. But this framing misses the structural problem.
AI systems do not crawl the web the way Google does. They are trained on datasets, fine-tuned on specific content, and augmented with retrieval systems that access particular sources. Visibility in AI requires different assets than visibility in traditional search:
- Structured data that AI systems can parse without interpretation
- Presence in training datasets that may be years old
- Integration with retrieval systems that AI tools actually query
- Brand mention frequency in sources the AI considers authoritative
Most hotels have none of these. They have websites optimized for human visitors and Google bots. They have OTA listings designed for comparison shopping. They have social media presence aimed at engagement metrics.
None of this translates automatically into AI visibility.
The uncomfortable truth is that many hotels cannot solve this problem individually. A 50-room independent property in a secondary market does not have the resources to build API integrations, generate machine-readable content at scale, or influence training datasets. For these properties, the path to AI visibility runs through intermediaries—DMOs, consortia, technology partners—who can aggregate their data and present it in AI-consumable formats.
This creates a new competitive dynamic. The intermediary layer that many predicted AI would disintermediate may instead become more important, not less. But the intermediaries that survive will be those providing AI accessibility, not human-facing marketing.
The Workforce Fracture
Skift reported this week on what they call "The Great AI Upskilling of the Travel Workforce." The picture that emerges is bifurcated. Some companies are investing heavily in training employees to work with AI tools. Others are reducing headcount and hoping the remaining staff figure it out.
According to the reporting, this gap creates risks both for individual companies and for the industry as a whole. Companies that underinvest in upskilling may find themselves unable to compete as AI capabilities advance. But the industry-wide effect may be more corrosive: a talent pool that becomes increasingly divided between AI-fluent professionals and those left behind.
Here is my contrarian position: the upskilling conversation is mostly focused on the wrong skills.
The dominant narrative is that travel professionals need to learn prompt engineering, understand large language model capabilities, and become comfortable with AI interfaces. This is true but insufficient.
The more valuable skill is judgment about when AI outputs are wrong.
AI systems are confident. They produce plausible-sounding responses even when those responses are fabricated or outdated. A travel professional who trusts AI outputs without verification is more dangerous than one who cannot use AI at all. The person who cannot use AI simply fails to capture efficiency gains. The person who trusts AI blindly makes confident errors at scale.
The upskilling that matters teaches professionals to use AI as a first draft, not a final answer. It develops intuition for the types of questions where AI excels versus where it confabulates. It builds workflows where AI handles volume while humans handle verification.
This is harder to teach than prompt engineering. It requires domain expertise, critical thinking, and comfort with uncertainty. But it is the skill that will separate high-value travel professionals from those replaced by AI—not because the AI is better at their job, but because the AI is faster at the parts of their job that were already routine.
The European Coordination
Travel And Tour World reported on tourism growth trajectories for Spain, Italy, Germany, France, and Portugal heading into 2026. The frame was familiar: expanded air connectivity, sustainability efforts, digital innovation. But the underlying story is about coordination.
European destinations are increasingly recognizing that they compete not just with each other but with destinations globally. And in an AI-mediated discovery environment, the destinations that present themselves coherently—with consistent data standards, integrated booking pathways, and machine-readable information—will outperform those that remain fragmented.
This creates an opening for DMOs willing to rethink their role. The traditional DMO function—marketing and promotion to human travelers—is being disrupted by AI. But a new function is emerging: data infrastructure for AI accessibility.
A DMO that aggregates structured data from all properties in its region, maintains it in machine-readable formats, and ensures it reaches AI training and retrieval systems is providing value that individual properties cannot replicate. This is not sexy work. It does not produce beautiful advertising campaigns or viral social media moments. But it may be the difference between a destination that appears in AI recommendations and one that does not.
The Hospitality Net Metric Shift
Hospitality Net published analysis this week on what they term "Agentic Hospitality"—the convergence of AI-driven discovery, integrated TravelOS platforms, and new tracking models from Google. The core argument is that traditional hotel metrics are becoming insufficient.
RevPAR, ADR, occupancy—these metrics assume a world where hotels control their distribution and can attribute bookings to specific channels. In an agentic world, attribution becomes murky. A traveler who books through an AI assistant may have been influenced by training data from years ago, real-time retrieval from multiple sources, and conversational context that is never logged.
This is not a measurement problem that better analytics can solve. It is a fundamental shift in how demand is generated and captured.
Hotels that thrive will be those comfortable operating with less attribution certainty. They will invest in brand building and data infrastructure not because they can measure the direct ROI, but because they understand that AI visibility compounds over time in ways that are difficult to track but impossible to ignore.
What This Week Actually Means
The common thread across these stories is not that AI is coming to travel. We knew that. The thread is that AI is becoming infrastructure, not application.
Airbnb is not using AI to improve their product. AI is building their product. Mindtrip and Sabre are not adding AI features to booking. AI is becoming the booking interface. Hotels are not competing for AI visibility. AI is determining whether they compete at all.
For DMO directors reading this: your marketing budget allocation for next year needs to include AI data infrastructure. Not as an experiment. As a line item comparable to your advertising spend. If your destination's properties are invisible to AI search, your advertising is reaching travelers who will never see your properties when they actually book.
For hoteliers: audit your machine-readable presence this quarter. Not your website SEO. Your structured data, your API accessibility, your presence in datasets that AI systems actually query. If you do not know where to start, that is itself the answer—you need partners who do.
For travel tech vendors: the companies that win the next decade will be those that serve as bridges between legacy travel inventory and AI consumption. This is unsexy middleware work. It is also where the value is migrating.
The game is not changing. The game has changed. The winners will be those who stopped optimizing for the old rules first.