When TikTok launched GO on May 12, the partner list was instructive: Booking.com, Expedia, Trip.com, Viator, GetYourGuide, Tiqets. Six companies that have spent years building machine-readable inventory pipelines. Airbnb was not among them—a deliberate absence that has generated considerable commentary about platform strategy.
That conversation is worth having. But it is obscuring a more uncomfortable one. The hotels debating whether TikTok GO matters to their distribution strategy should first ask: Is their inventory structured in a way that any of these systems can actually read?
For a significant portion of the industry, the answer is no. And the implications of that gap are arriving faster than most properties are prepared for.
Booking.com did not join TikTok GO on launch day by accident. They began testing hotel sales inside TikTok in August 2025, starting with roughly 10% of US users—months before the formal launch.
That testing window was not primarily about conversion optimization. It was infrastructure validation: confirming that their inventory pipeline could feed a new channel at scale without rebuilding from scratch. The test succeeded because Booking.com's entire architecture is built around one operating principle. Inventory must be readable by any system, through any interface, at any time.
This is not a technology advantage in the conventional sense. It is a strategic posture that Booking.com adopted years ago and has been compounding ever since. Every new channel—metasearch, voice and now in-app social commerce—is available to them within weeks because the underlying data architecture was designed for exactly this kind of interoperability.
Airbnb's calculation is different and worth understanding on its own terms. Brian Chesky has argued publicly that chat-based interfaces are architecturally wrong for travel—too text-heavy, incapable of direct manipulation, poor at comparison across thousands of options and fundamentally single-player in a category where most bookings involve multiple people. Airbnb is not sitting out TikTok GO because they missed the meeting.
They are sitting it out because they are building what they believe is a better interface than the ones currently available, with Ahmad Al-Dahle—formerly head of generative AI at Meta and architect of the Llama model family—now serving as CTO. Joining a competitor's distribution infrastructure while racing to make that infrastructure obsolete would be a strategic contradiction.
Both positions are coherent. The hotels stuck in the middle of this debate, however, face a different problem entirely.
The French experience after 2014 is the closest historical parallel and it remains underappreciated. When French regulators removed Booking.com's wide price parity clauses, hotels could finally offer better rates on their own websites. The expected shift toward direct booking did not materialize. A 2022 European Commission market study found that 79% of hotels still did not differentiate prices between OTAs, and the competitive landscape showed minimal change from 2016 despite years of regulatory intervention.
The lesson was not that direct booking is impossible. It was that price was never the primary reason guests were booking through OTAs. The reason was architecture—the speed, familiarity and frictionless experience that Booking.com had built into its discovery-to-transaction flow.
Hotels that responded to the parity clause removal by adjusting their rates were solving the wrong problem.
TikTok GO is that same architecture problem, accelerating. The booking journey is moving further from the hotel's control not because travelers prefer intermediaries but because the infrastructure of inspiration, trust and seamless payment is being built somewhere else, by someone else, at a pace most properties cannot match through incremental website improvements.
This is where two technical concepts move from background noise to operational priority.
JSON-LD—JavaScript Object Notation for Linked Data—is the structured data format that search engines, AI assistants and platforms like TikTok GO use to understand what a hotel is, what it offers, what it costs and whether it is available on a given date.
A hotel without properly implemented JSON-LD is not penalized in the next generation of travel search. It is simply absent. The algorithm has no legible signal to work with and moves on to properties that provide one.
Most hotel websites were not built with machine readers in mind. They were designed for human visitors—visually compelling, navigationally intuitive, optimized for a user who scrolls and clicks.
An AI agent does not scroll. It queries structured data, extracts what it needs and either acts or moves on. That gap between human-readable and machine-readable is already a distribution gap. It will become a significant competitive one.
Model context protocol (MCP) takes this further. Where JSON-LD makes a hotel's public-facing data readable to systems that visit the website, MCP enables AI agents to connect directly to inventory systems—retrieving real-time availability and pricing, and in some implementations completing bookings, without any website visit at all.
HomeToGo has already launched an MCP server and a live ChatGPT integration. These are not experiments—they are infrastructure investments in a distribution channel that does not yet have significant volume but is being built now, before the volume arrives.
There is a third layer that receives even less attention: the internal knowledge infrastructure that hotel operations run on.
Standard operating procedures, service protocols, rate plans, property specifications—the documents that govern how a hotel functions—exist in most properties as a collection of Word files, PDFs and institutional memory scattered across shared drives.
Vector databases that index this documentation and make it queryable by AI agents in real time are the mechanism by which any property can begin to operationalize AI tools without relying on staff to manually locate and interpret information under pressure. The external and internal infrastructure problems are the same problem viewed from different sides of the front desk.
Independent hotel reliance on online travel agencies (OTAs) increased in 2025, with OTA share reaching 63.4% of bookings, up from 61.3% in 2024. The direction of travel is not ambiguous.
What is striking about this moment is not that hotels are behind—that has been true for years—but that the distance between where the infrastructure conversation is happening and where most properties are operating is widening at exactly the point when the decisions made now will determine distribution positioning for the next decade.
Booking.com spent months validating their TikTok GO pipeline before launch. HomeToGo built an MCP server before MCP became a mainstream distribution channel. The pattern in travel technology is consistent: The infrastructure work happens quietly, before the volume, and by the time the volume arrives the window for catching up has already narrowed considerably.
Hotels cannot compete for AI-driven traffic by adjusting their rates. The French experiment demonstrated that a decade ago. What they can control is whether their inventory is legible to the systems increasingly mediating between traveler intent and accommodation choice—externally through JSON-LD and MCP, internally through structured knowledge infrastructure that makes AI genuinely useful to the people running the property.
The question is not whether TikTok GO matters. It does. The question is whether a hotel's current infrastructure would let them be part of whatever comes after it.
About the author...
Boštjan Koželj has worked in hospitality IT for over a decade and is completing doctoral research on digital platforms and tourism.