Hotels Are About to Lose Another Distribution War. This Time to AI Assistants.

Source: Henry Ly; Predict (Medium) · 2026-04-07

A new protocol called MCP lets AI book rooms directly. Most hotels don’t know it exists yet.

This is not a concept demo. Perplexity already processes end-to-end hotel bookings through its Tripadvisor and Selfbook partnerships. Google has committed to transactional trip planning inside Gemini and AI Mode, backed by partnerships with Booking.com, Expedia, Marriott, and IHG (Skift, 2025). OpenAI is opening its app marketplace to travel companies beyond the current OTA partners. The protocol that makes all of this possible is called MCP, Model Context Protocol. And most hotels have never heard of it.

MCP eliminates this. A hotel builds one MCP server that wraps its existing systems in standardized tool definitions. Any AI assistant that supports the protocol can connect to it. One integration, every AI. As Google Cloud describes it, MCP allows AI tools to move beyond static knowledge and become dynamic agents that can retrieve current information and take action.

OTAs take 15–25% per booking. MCP takes zero.

The economics are what make this existential for the hotel industry. OTAs currently hold approximately 55% of hotel booking market share. Their commissions range from 15% to 25% per booking. On a $200 room, that is $30 to $50 going to the intermediary every night.

But the distribution landscape is shifting again. A 2026 survey by TakeUp found that 90% of US travelers are already aware of AI travel assistants. The travelers who adopt AI-assisted booking will not compare your website against Booking.com. They will ask their AI assistant to find the best option, and the assistant will query whatever inventory it can access through MCP.

Hotels that have deployed MCP endpoints will appear in those results with live rates, real availability, and the ability to complete the transaction. Hotels that have not will be represented by whatever third-party data the AI can scrape, which is often inaccurate, outdated, or filtered through an OTA that takes a commission anyway.

RateGain launched the first MCP-enabled booking engine in 2025. Most hotels missed it.

The industry is not waiting for consensus. RateGain Travel Technologies launched the first MCP-enabled booking engine in late 2025, allowing AI assistants to transact directly with travel inventory. Sabre, PayPal, and MindTrip announced a partnership in February 2026 to build an end-to-end agentic booking pipeline: a traveler describes their trip in natural language, MindTrip queries Sabre’s Mosaic APIs (420+ airlines, 2 million hotel properties), and PayPal handles payment within the conversational flow. The entire loop completes without the traveler opening a browser. Planned launch: Q2 2026 (OAG, 2026).

citizenM’s CIO Mike Rawson captured the operational shift precisely: MCP moves hotels “from coding to describing intent.” Instead of requiring developer teams to build and maintain API integrations, hotel operators describe what they want to expose (availability, rates, policies, amenities) and the MCP layer translates that into tool definitions that AI assistants can use.

This matters because most independent hotels and small chains do not have engineering teams. They rely on their PMS vendor or booking engine provider to handle distribution technology. If those vendors do not add MCP support, the properties they serve become invisible to AI-driven booking.

The critical point is that MCP does not replace existing hotel systems. It sits alongside them as an orchestration layer. The PMS, CRS, and booking engine continue to function as systems of record. MCP simply provides a standardized way for AI assistants to read from and write to those systems. This is the foundation of what the industry now calls agentic AI in travel: AI that does not just answer questions but takes autonomous action on behalf of travelers.

The parallel to mobile optimization in 2012 is instructive. When smartphones became the primary browsing device, hotels that optimized their mobile experience early captured disproportionate direct traffic. Hotels that waited saw OTAs fill the gap with superior mobile booking interfaces, cementing the intermediary’s dominance for a decade.

MCP represents the same inflection point for AI-driven distribution. By the end of 2026, MCP-enabled travel bookings will be live across multiple AI platforms, not theoretical, not demo-only, but real transactions where an AI agent queries a hotel’s availability, rates, and policies directly through MCP and completes the booking (The Paypers, 2026).

Hotels that deploy MCP endpoints now will establish themselves in this channel before it becomes crowded. Hotels that wait will find themselves in the same position they occupied with mobile: dependent on intermediaries who moved faster.

The hotels that understood the OTA threat in 2010 invested in direct booking technology. The hotels that understand the AI assistant threat in 2026 will invest in MCP. The question is whether the industry will respond faster this time, or whether a new class of intermediaries will capture the channel before hotels realize it exists.

Travel software development companies that already build custom booking engines and GDS integrations are now adding MCP server endpoints to their delivery scope. The technology is not the bottleneck. The bottleneck is awareness. And right now, most hotels do not know MCP exists.