What artificial intelligence is really changing in travel agencies, hotels and tour operators. No hype, no brochure promises.
MCP (Model Context Protocol) is an open protocol connecting AI models to external systems — booking engines, PMSs, databases — and in 2026 it consolidated as travel's de facto standard: adopters range from Sabre to SiteMinder. It solves the underlying problem: a hotel's website exposes only 5-10% of its operational information, and assistants need live data to answer with certainty.
Read the articleBooking closed 2025 with record revenue of 22.82 billion euros (+13.4%) and a net profit 8% lower (4.58 billion), according to its official results. The causes: 770 million extra in marketing — mostly Google — and a 388-million write-down of its metasearch engine Kayak, damaged by AI search. The OTA that charges hotels commissions pays, in turn, a structural toll to Google, just as Google starts competing with it.
Read the articleThe practical difference between a chatbot and an AI agent is that the chatbot tells you how to book while the agent books: it sets goals, plans and executes multi-step tasks with little supervision. Spain already has cases with figures — RIU's assistant handles 75% of more than 1,500 daily queries — and a three-level ladder lets you place any system in minutes.
Read the articleHotel AI cases with results verified by independent sources can be counted on one hand: Wyndham (−25% handling time across its AI toolset, quarterly results, 2025), Hyatt (almost +20% group sales productivity, stated in its own 2026 results) and Hotel Tech Report's revenue benchmarks (+15-20% RevPAR, across more than 4,800 hoteliers). Almost everything else is vendor data: quotable, but with a label.
Read the articleThere is no stable 'position' in AI answers: across 13,500 searches, the top recommended hotel changed in 45% of same-day repetitions of the same query, and the five big platforms agree on the number one only 4% of the time (Kollective, July 2026). The useful metric is not a point-in-time position but consistency of appearance per platform over time.
Read the articleBetween January 2025 and August 2026, AI travel booking went from demo to reality: Booking and Expedia apps inside ChatGPT, the first flight booked entirely in a chat (Mindtrip with Sabre and PayPal, May 2026), OpenAI's checkout retreat and a Google hotel booking test in AI Mode. The lesson of so many twists: the real battle is not who charges for the sale, but who controls discovery — and OpenAI is already monetising it with ads.
Read the articleA quick guide to the 18 terms dominating the AI-and-tourism conversation in 2026: from fundamentals (LLM, RAG, hallucination) to assistant visibility (AEO, GEO, AXO, llms.txt) and agentic booking (agent, MCP, zero-click booking). Each term explained in plain language, with why it matters to a travel business.
Read the articleDocumented AI-driven job cuts in tourism have so far hit corporate and support roles — Accor expects AI to replace up to 40% of its office jobs, Mews laid off 15% of staff — while customer-facing employment holds up: travel agent was Spain's fastest-growing profession according to LinkedIn (2025), though it no longer appears on the 2026 list. There is no single narrative: there are three at once.
Read the articleA travel business's visibility in AI assistants works in three layers: the model's memory (what it absorbed during training), live web search, and dynamic sources connected in real time. Each layer has different rules and degrees of control, and confusing them leads to investing in the wrong place.
Read the articleAround 90% of consumers want to plan trips with AI, but only 2% to 7% would let it book and pay autonomously, as Booking.com, Skift, McKinsey and Accenture all found in 2026. The gap is not technological but one of trust: AI still cannot guarantee verified data or answer for mistakes when something goes wrong.
Read the articleAI assistant crawlers don't execute JavaScript, agents give up when they hit captchas and long forms, and a hotel's website typically exposes only 5-10% of its real operational information. This 7-point checklist lets you verify in one afternoon whether your site is readable — and usable — by the AIs that already recommend and are starting to book travel.
Read the articleAt independent Spanish hotels, 76.7% of bookings arrive through OTAs — only Portugal (79.7%) and Indonesia (77.9%) depend more, according to the Cloudbeds report based on 90 million reservations (2025). Spain also leads the world in OTA cancellations at 26.6%, more than double the direct channel's rate.
Read the articleWhen ChatGPT, Gemini or Perplexity recommend a hotel, more than half of their citations come from OTAs (55.3%) and only 13.6% from official hotel websites, according to a Cloudbeds study based on 810 queries. Recent reputation and third-party presence weigh more than your own website.
Read the article
