Generative search is probably the biggest disruption in hotel digital marketing since mobile internet. It’s moving fast, and it has a long way to go. Say goodbye to old habits. Welcome to algorithms which inspire, plan, and book. So what now for your good old hotel website?
Google was yesterday. Today, generative AIs are constantly whispering advice in our ear. Assistants like ChatGPT Operator, Google Gemini, or Microsoft Copilot are turning into true digital concierges. But where does their data come from?
Firstly, it comes from the web itself: indexing everything they can find. This is costly in time and energy, thus expensive and definitely not very efficient.
Secondly, to get fresher data, AIs will most likely rely on the OTAs (Booking, Expedia…). This is logical because they hold the largest catalogue. However the issue is that the OTAs only list the hotels they have agreements with and they only show the rooms these hotels allocate to them, and sometimes at higher rates (yes, direct booking can often be cheaper). All in all not ideal for an AI agent which is supposed to display the full picture. And what about website visits ? Perplexity now uses Comet, an intelligent browser that can plan, compare, and book a hotel without ever actually visiting the property’s website.
As a result, hotels can be completely bypassed — no more search, no more visits, no more direct bookings. The AI decides which hotel gets chosen.
Comet illustrates the rise of agentic AIs, which act on behalf of the traveler. The guest simply says, “Book me a hotel in New York,” and the AI selects and confirms the reservation automatically. This is no longer a technical issue — it’s a commercial shift: the customer journey now happens inside the AI systems.
To stay visible, hotels must structure their data and train their own AIs. This will lead to the third phase — the era of the Model Context Protocol (MCP): imagine a box where hotels put all their information (content, images, availability, rates, CRM etc.), enabling AIs to fetch the data they need in real time, to provide travelers with access to everything inside their interface. In other words, like a mini-website generated instantly from the initial request (content, photos, availability, rates). Clearly, if everything is up-to-date and accessible in real time in the MCP, the traditional hotel website loses much of its efficacy.
In theory, thanks to these MCPs providing better-quality information, hotels could completely bypass the OTAs. Picture this: you ask your AI assistant to book a room. Using the MCP, the agent checks rates, applies a loyalty programme, processes the payment, and confirms the booking all without even visiting a website or touching an OTA.
That said, let’s not bury the OTAs just yet. They’re resilient, and they still offer undeniable advantages (trust, assistance, re-accommodation, free nights, installment payments, cashback, etc.).
Some say hotel websites are doomed: moving forward AI will be able to handle everything, from information, details and photos to rates and bookings. However today, only 25% of AI answers come from official content; the rest comes from reviews, blogs, travel agencies, social media, etc. This mix of unstructured data sources explains why the information provided isn’t always reliable, and as we all know, sometimes it is completely invented – this is the infamous “AI hallucinations.” (Please see the article: Failure Is Not an Option, It’s a Certainty).
In reality, today, hotel websites do remain central: ultimately they are the official source that people trust, the space where hotels display their full data, tell their story and deploy their own AI (like Velma). The website provides certified information, showcases the latest photos/ events/updates, and expresses the hotels personality.
Tomorrows websites however, will need to be built for both human narrative, and be interactive and engaging for AI, through structured content and tailored FAQs. They’ll need to become dynamic: pages will need to adapt not only to criteria like IP address, browser, language, date, and time, but also to the customer’s profile (for example, loyalty memberships) and even the customers’ conversations with the site’s own AI.
Tell your story, highlight what makes you unique: local history, atmosphere, details found nowhere else. AI thrives on semantic richness, not on copy-paste.
Think not only SEO but also AEO (Answer Engine Optimization): structured, conversational, authentic content aligned with travelers’ real questions. Think Dynamic FAQs by language and by season (people don’t ask the same in summer and winter). Tools like Q-SEO are without a doubt the most effective.
Take care of your internal data (first-party data): thorough descriptions, amenities, reviews, local partnerships… so the AI won’t need to invent or make things up.
Spread these across every channel: website, blogs, social media, influencers, guides… The wider your presence, the more AIs will be able to pull directly from you. (See the article: Is SEO Dead?).
Imagine a traveler asking their AI:
“Find me a hotel in Brooklyn NY, near Williamsburg, with a safe big enough for a 15” MacBook Pro, USB ports in the room, an ice machine, and offering a gluten-free breakfast, for three nights starting October 20, at under $270 a night.”
In a world of AI agents, hoteliers will need to transition swiftly from their official website, to their official FAQ and their official database.
This is where the real battle will take place – with transactional queries with high intent like this one – this is the key to direct customer engagement. No website today can provide all of this information exhaustively.
So what is the solution?
A structured, comprehensive database which collects and collates every detail, and one which connects to large language models through the MCP’s (see first section).
To put it simply: we’ve already moved from the official website to the official FAQ. The next step will be the official database (like Q-Data), accessible through the hotel’s MCP, ensuring total precision and maximum depth. (See article: From poor data quality you will suffer…).
Today, a hotel is estimated to be defined by around 4,000 data points but a typical hotel website only holds around 200. Even Booking.com only holds about 400. Q-Data, the official database for hotels, contains an astonishing 3,100. It’s easy to see: the more structured data a hotel owns, the more effective it will be in the AI-driven world.
Generative search and AI agents can give hotels a unique opportunity to take back control of their brand, their data, and their guest relationships. Essentially the key to this success will be combining the technology, the official database and the connectivity with the experience, and the human touch.
A word of caution, though: searching and booking will most likely remain as two separate stages. Firstly we can turn to an AI agent to handle the accommodation search. However, when it comes to finalizing the reservation, it’s likely that we will continue to rely on the trusted channels we already know, such as the hotel’s website, the OTAs, or perhaps even the good old telephone.
© Image: Shutterstock
Learn the 10 proven methods to increase direct bookings and reduce third-party dependency.
Your website gets traffic, but are they booking? Engage visitors instantly with Velma answer questions 24/7, and boost direct revenue.
Download the free checklist 
Convert more visitors today 
