Direct selling is being shaped by artificial intelligence: reasons why MCP has become the new competitive landscape.
Over the years, the struggle for direct sales has been marked by familiar elements: Google, online travel agencies, search engine optimization, search engine advertising, metasearch engines, the hotel website and the reservation system. Although the circumstances varied, the goal remained the same: to capture clicks and transform them into direct bookings.
By the year 2025, that model has become obsolete without causing much of a stir. The traveler has stopped "searching" for information and started asking questions. Now, the answers come not just from Google, but from a new set of conversational assistants such as ChatGPT, Gemini, Perplexity and Claude. These do not offer a list of ten links, but provide a single answer, an integrated application or even a quasi-confirmed booking.
In TecnoHotel we have addressed in several articles the issue of the disappearance of the click, the competition between Google AI Mode and ChatGPT to attract travelers, as well as the incorporation of ChatGPT as a new distribution channel, with OTAs being the first to participate.
In this context, a new layer is added that you are probably familiar with: the MCP (Model Context Protocol), a standard that has the potential to determine whether the direct channel is closed or transformed again.
From interaction to dialogue: the tourist no longer accesses your website
Toni Mascaró mentioned this during a spontaneous discussion at TIS: "Artificial intelligence has become an ally of the traveler. There is no such thing as a click, as the search is done through platforms such as ChatGPT, Perplexity or Gemini". The traveler expresses his wishes and the AI responds with a comprehensive proposal, which can be modified through conversation, without the need to leave the assistant.
Google AI Mode applies this approach to search: when a user requests "a spa hotel in Lisbon for two nights that accepts pets and includes transportation", the response includes a map, prices and several options directly in the Google interface. On the other hand, ChatGPT incorporates applications such as Expedia, Booking or Kayak into the chat (although for the moment only outside the European Union): the user asks for ideas, the system identifies the intention to make a reservation and suggests opening the OTA application, which appears within the conversation.
In both cases the same thing happens:
This article about Google AI Mode warned that the number of clicks directing to official websites will continue to decline; much of the discovery and decision making process is done directly in the search engine or assistant interface. In short, OTAs are once again overtaking hotels, but this time in the chat realm.
The situation for the direct channel is clear: if conversations have become the new showcase and your hotel doesn't appear as an understandable source for artificial intelligence, then it practically doesn't exist.
What role does MCP play in all this?
MCP may look like just another set of acronyms, but it's really not. It is an open protocol being used by companies such as Anthropic, OpenAI, Google and Microsoft. Its goal is to enable artificial intelligence agents to communicate in a standardized way with external systems such as databases, CRM, booking engines and ERP.
In hotel industry parlance, MCP acts as a universal adapter that facilitates communication between an artificial intelligence agent and your property management system (PMS), central reservation system (CRS) or booking engine, without the need to create a new integration each time.
The logic is simpler than it appears.
The advantage of MCP lies not in being a current trend, but in the following: if your vendor implements MCP once, any compatible agent (such as ChatGPT, Claude, Gemini, among others) can be linked without the need to restart the integration from the beginning.
It is the breakthrough that could make it possible, in the future, for a traveler to book directly into your inventory through an artificial intelligence assistant, without having to use Booking.
That "someday" doesn't happen by magic. It is the result of careful planning.
Three emerging approaches to hotel sales using artificial intelligence
From a hotel manager's perspective, the picture is shaping up as follows:
OTA model: the official application available within ChatGPT. It is the one that is currently operational outside the European Union.
Benefit: by making an effort with that OTA, you will gain greater exposure in the new channel. Drawback: at the same time, you strengthen its role as an intermediary in the age of artificial intelligence and still pay the usual commission.
2. CCM aggregation model: the new "meta-mediation".
New entrants are emerging that pitch themselves as, "Join my platform and I will provide visibility into all AI agents that support MCP."
Sounds innovative. However, the business model remains traditional: a commission in exchange for exposure. Simply put, MCP aggregators, such as DirectBooker, are emerging that act as intermediaries between hotels and artificial intelligence assistants. These aggregators receive inventory directly from multiple hotels, organize it and present it to platforms such as ChatGPT, Gemini or Claude from a single MCP server.
Technological innovation, but with familiar principles: another cost between hotel and customer, this time in the form of procedures. If the industry settles for this, it may fall into the same pattern as before: a new dependency, this time driven by artificial intelligence.
3. "MCP-first direct" model: the one that really matters to the hotel... if your supplier is bold.
This is the scenario that truly transforms the situation:
At present, it is not widespread, but the technical possibility exists. What is required is not more technology, but a firm decision on the part of the suppliers and pressure exerted by the hotels.
Mirai has moved in this direction by presenting its proposed MCP server, which is based on a canonical database to manage content and reservations, as Pablo Delgado explains in his analysis.
Neobookings has decided to follow this direction, focusing on its engine working as a direct access between hotels and artificial intelligence assistants, avoiding the intermediation of OTAs or aggregators. Moreover, Paraty Tech is already up and running, as David Madrigal mentioned a few days ago.
Will there be direct bookings through ChatGPT? Yes. The question is: from whom?
Visualize a situation two years from now:
- "I need a hotel in Valencia for two nights, including breakfast, with a large bed and that accepts pets. My budget is up to 140 euros." - The wizard offers four alternatives, combining online booking sites and direct options. - The customer selects one: "Confirm the third option, which has flexible cancellation." - The wizard replies, "Perfect, your booking has been made. You will receive an email from the hotel."
In the background, two situations may have occurred:
From the traveler's perspective, the experience is identical. However, from the hotel's perspective, it is completely different.
The distribution of that future is being defined at this time, while there are those who continue to state: "When this is more advanced, then we will see".
SSOT, reservation system and content management system
Pablo Delgado's work in Mirai is very relevant in this context, as it organizes a concept that many recognize but few know how to express: before addressing CCM, hotels face a more fundamental challenge: chaos in information management.
His suggestion is to create a single, definitive database (Single Source of Truth, SSOT) containing content and booking information, such as: room types and features; amenities, hours, regulations and restrictions; dining, spa, activity and kids' club options; guest FAQs; availability, pricing, promotions and commercial policies; and loyalty program levels and benefits.
This "singular truth" is what should nourish the CCM server. Likewise, Pablo presents a very defined distribution of functions.
His conclusion aligns with what you have been arguing in TecnoHotel: the best way for a hotel to communicate directly with artificial intelligence is through the combination of a booking engine (for all transactions), a supplier offering full content and the MCP as the standard connection protocol.
The forbidden word: "autodiscovery".
In most technology circles, the term autodiscovery is gaining popularity: it refers to the concept that, instead of relying on a limited catalog of applications, assistants are able to automatically identify which MCP servers are available (such as those in hotels) and join them.
Is there currently a global mechanism of this type? No. Is this a logical evolution? Yes.
On the other hand, MCP aggregation layers, such as DirectBooker, are already available that anticipate the future of autodiscovery and position themselves as a 'meta' between your inventory and artificial intelligence assistants.
The main goal would be for artificial intelligence to "monitor" which hotels have made a good preparation (with their MCP and SSOT in correct condition) and offer them as direct sources, without the need to resort to an OTA or an aggregator.
However, it is important to be clear: there is currently no official date or definitive global standard for autodiscovery. Having an MCP server does not guarantee an instant connection to artificial intelligences, but it appears that, if MCP becomes firmly established, autodiscovery layers will eventually be implemented.
When this happens, hotels with a functioning MCP server will have a significant advantage. The others, on the other hand, will have no place in that type of market.
MCP-ready: superficial etiquette or authentic commitment
In this scenario, the "MCP-ready" qualification could become a mere fad... or a major issue.
A technology provider that, by 2025-2026, wants to support this claim should be able to present at least one MCP server (even if it is in beta) that allows verification of availability, rates and basic policies. In addition, it would have to have a security and permissions model designed for artificial intelligence, specifying who has access to what information, from what location and for what purpose. It should also perform concrete tests with at least one agent, and not just present attractive PowerPoint presentations.
Anything else is just distractions. In addition, the hotel has full authority to ask the tough question, "Are you arranging a direct channel for me or are you going to leave me dependent on OTAs?"
If the answer is the second option, you do not have a direct channel partner, but you have an additional intermediary.
Final analysis: the true situation of direct sales
All in all, the picture is clear:
First, a change of perspective is needed: abandon the idea that artificial intelligence is just a "passing trend" and accept that it has now become the new form of interaction for travelers.
Direct selling still exists; what is really in danger is the belief that it can be sustained only with an attractive website, good search engine positioning, effective advertising campaigns and a good platform.
As interactions with artificial intelligence become a fundamental part of travel, competition will shift from focusing solely on capturing clicks to focusing on being present in the sources that AI can query and employ.
In this context, MCP and canonical databases are not mere technical terms; they are the components that determine who gets a chance to play in the major league... and who doesn't even make it into the game. The upside is that there is still time to change the outcome. The negative is that the game has already started.
If the hotel industry wants a traveler in 2027 to be able to say "book directly at hotel X" through a virtual assistant, and actually have the booking go to the hotel system and not to an online travel agency, it is crucial that in 2026 they focus on concrete developments and not just on attractive presentations.
Artificial intelligence will not worry about figuring out who was late, it will simply use what is available.
Notice:
In the next few days, we will be launching a checklist with 15 questions that you can use to meet with your supplier to see if they are really ready for MCP, or if they only mention it in their presentations.
Images provided: Image created by artificial intelligence.
Authors
Director of TecnoHotel. He holds a degree in Journalism from the Complutense University and has more than a decade of experience in the educational and cultural field. Since 2017, he has been in charge of the editorial and coordination of TecnoHotel. He has held roles as editor and editor-in-chief, and was appointed director in September 2023.
