AI & Travel Distribution: ChatGPT, Google, and the Battle of New Standards (UCP, MCP, GEO)
AI & Travel Distribution ChatGPT, Google, and the Battle of New Standards (UCP, MCP, GEO) In less than 18 months, travel distribution has entered a new phase. AI assistants are no longer just advising; they are beginning to execute, plan, compare, and book. This shift toward a « clickless » web is reshuffling the cards of the value chain: – Who controls the customer interface? – Who provides the « product truth »? – And which technical tracks—protocols, standards, content—will bookings now run on? Signals are multiplying: the creation of native Apps within ChatGPT by both incumbents and newcomers, AI Mode in Google Search, the Universal Commerce Protocol (UCP) for agentic commerce, the Model Context Protocol (MCP) to connect AI to legacy systems, and the rise of GEO (Generative Engine Optimization) to maintain visibility when AI answers in place of websites. This overview of new standards draws on a series of use cases illustrating how these technologies are permanently transforming the travel value chain. From « Search & Click » to « Ask & Act » The historical paradigm of digital travel was linear: inspire → search → click → compare → book. Generative AI breaks this model. When an assistant synthesizes, recommends, and orchestrates within a single response, it captures a growing share of intent upstream of the click. This evolution is already fueling concerns among content publishers facing the rise of AI-generated answers and the gradual disappearance of discovery traffic (as noted by The Guardian). In travel, the impact is even more structural. Booking relies on complex transactional objects: real-time availability, pricing rules, cancellations, and servicing. It is precisely in this context that agentic AI—capable of taking action, not just providing answers—becomes a new interface operator. As one industry expert puts it: « AI can only function if it has reliable data… but our industry is extremely fragmented. » The promise is significant, but the inherent complexity of travel remains. Content Exhaustiveness and Personalized Recommendations: The New Equation Since late 2025, OpenAI has taken a strategic turn by opening ChatGPT to native applications integrated via its marketplace (Apps SDK). ChatGPT is no longer just a conversational assistant; it is becoming a direct service access platform. In the travel sector, early partners include Expedia and Booking.com, quickly joined by Tripadvisor, with plans to extend to other stages of the journey (mobility, experiences, dining). Expedia fully embraces this strategy: « We want to meet travelers where they are. Integration into ChatGPT is a major step. » What this changes concretely: Before: ChatGPT → OTA site or app → search → filters → booking. Now: The user expresses their need in natural language; ChatGPT queries the application directly (inventory, availability, sorting, recommendations) and brings decision and action together within the same conversational space. Implications for distribution: New entry point for demand: ChatGPT can become the upstream homepage, capturing intent before SEO, websites, or apps. New competition for data: The advantage shifts toward players capable of providing rich inventories, real-time actionable pricing, and executable rules. Reconfiguration of intermediation: If the interface shifts to the assistant, value focuses less on the front-end UX and more on access to supply, brand trust, and servicing capabilities. Google: From Planning to Booking Directly in Search Google is following a parallel trajectory, this time at the very heart of its Search engine. In November 2025, the group introduced travel features integrated into AI Mode, aiming to unify planning and action. Key building blocks include: – Canvas: A planning workspace (flights, hotels, Maps data). – AI-powered Flight Deals. – Agentic capabilities allowing users to turn a plan into a booking via partners. PhocusWire confirms this direction: Google clearly aims to extend flight and hotel bookings, positioning AI Mode as a central hub of the traveler journey. The CEO of Kayak summarizes the impact on acquisition bluntly: « Organic search is dying. » The key takeaway: Google isn’t starting from scratch. With Flights, Hotels, Things to Do, and Maps, AI acts as an orchestration layer: capturing intent, structuring relevant options, and accelerating the path to purchase without breaking the user experience. UCP: Google’s Bet on an Agentic Commerce Standard… Applicable to Travel? In January 2026, Google announced the Universal Commerce Protocol (UCP), an open-source standard designed to structure agentic commerce from end to end—from discovery to post-purchase. The goal is to provide a common language between AI agents, merchant platforms, and transactional blocks. On paper, UCP could streamline a booking journey entirely driven by AI. In practice, travel is resistant. Travel products involve dynamic pricing, volatile inventory, ancillaries, multi-segment scenarios, and high post-sale intensity. PhocusWire nuances the immediate impact of UCP: Google might first capitalize on its existing integrations to solve the « first mile » (intent and qualification) before broader standardization takes hold. MCP: Connecting AI Assistants to Core Travel Systems The Model Context Protocol (MCP) addresses a complementary challenge: connecting assistants to the systems where data and business functions actually reside. MCP standardizes how an application exposes to an AI: – Reliable context (data, rules, constraints). – Executable functions (search, book, modify, cancel). It could become the technical bridge between conversational agents and the travel stack: PMS, CRS, booking engines, inventories, GDS, CRM, and loyalty programs. Initiatives attributed to players like Kiwi.com or Sabre illustrate the industry’s interest, even if implementations remain heterogeneous. Kiwi.com already sees it as a foundational standard: « MCP is becoming the preferred interface for AI agents to discover and use services like search or booking. » Governance: A point of vigilance. Connecting AI agents to transactional systems expands the risk surface (rights, security, injections). Vulnerabilities observed on some MCP servers serve as a reminder that no protocol can be deployed without robust governance. GEO: When Visibility is Won in the AI Response, Not on Your Site With generative AI, ranking well is no longer enough. The challenge is now to be cited and synthesized within the AI’s response: this is the shift from SEO (Search Engine Optimization) to GEO (Generative Engine Optimization). For travel players, the dilemma is clear: – Gain visibility via AI channels, – Without





