Asking whether a hotel accepts pets and booking a room are different tasks, although a traveller may consider them part of the same conversation. The first needs a reliable answer about the property; the second requires availability, prices, conditions and an operation completed in a reservation system. Mirai has begun connecting both dimensions with ChatGPT so its clients' direct channel can be used through the assistant as well as through the hotel's website.
The company announced the deployment on September 23, 2026, identifying the Emperador and Moderno hotels in Madrid, Porcel Hoteles in Granada and Gloria Thalasso & Hotels in the Canary Islands. According to Mirai, users can consult information and book directly with those brands using natural language. The announcement places the new commercial entry point at two individual hotels and two chains, with deployments that the provider will keep limited while it learns from their use.
The proposition addresses a specific hotel-distribution problem. A conversation can guide a guest, but it also needs to reach the offer a hotel can sell at that moment. Mirai connects enquiries with its booking engine's capabilities: searching for dates and occupancy, comparing rooms and rates, applying commercial conditions and acting on a reservation. September's development brings that infrastructure into ChatGPT, where the company describes its connections as plugins.
A traveller's question contains a commercial operation
Mirai's example begins with an ordinary request: a family trip to Madrid, a room with breakfast and a cancellable rate. Every element adds a condition to the search. Recognising the destination or providing an attractive photograph is insufficient. A useful response must connect occupancy, dates, product and commercial policy to present options the traveller can purchase. The assistant therefore needs both information about the property and access to the direct channel's functions.
Mirai distinguishes knowledge from capabilities. Knowing whether pets are accepted belongs to the first group. Checking availability for two adults and a child on specified dates belongs to the second. A rate has a similar distinction: explaining its conditions and quoting a price are different actions. The integration allows those consecutive questions to use the hotel's information and rules, rather than treating the dialogue as separate from what the engine can sell.
The announcement's described functions include checking prices, changing currency, filtering results and exploring promotions and rooms. They also include creating a reservation and modifying or cancelling an existing one. The commercial relationship can consequently continue after a stay is selected: the same conversational access is intended to help manage the operation. A visitor expresses a requirement and the technology translates it into enquiries or actions in the system administering direct sales.
This makes a meaningful difference for marketing. The destination of a contact need not always be a page where the user has to enter the request again through selectors. A conversation can retain the context used to compare options and progress towards a booking. What the channel gains is another way to use product, conditions and availability, alongside a new place to present information about the hotel brand and respond to a potential guest.
The connection rests on an earlier architecture
Mirai introduced its conversational infrastructure on March 12, 2026. That precedent helps explain the current deployment: the company had worked on the engine and the components bringing together content, business rules, conversation and external connectivity. September's ChatGPT connection rests on that core. The company describes it as an entry point into existing capabilities rather than the creation of a second inventory held separately for the assistant.
The architecture has five components. The engine processes quotes and reservations; Knowledge organises structured hotel content; Sarai conducts the conversation; Lobby supports supervision of activity; and MCP connects external assistants. Each has a role in the commercial journey. Information about services and policies meets inventory and sales rules so a request can become a proposed stay and subsequently an operation, with the components working around the same booking system.
| Component | Function described by Mirai |
|---|---|
| Booking engine | Quotes, inventory, commercial rules and reservation operations |
| Knowledge | Structured hotel content and knowledge |
| Sarai | Conversational interface with the guest |
| Lobby | Supervision, leads, measurement and human intervention |
| MCP | Connection of capabilities with external assistants |
Collapse notes
MCP, Model Context Protocol, is the layer Mirai uses to expose tools to external systems. In this project, those tools make it possible to consult information, search availability and prices and act on reservations. The company was developing this approach before its announced ChatGPT connection: its explanation of MCP dates from December 2025. That chronology connects preparation of transactional capabilities with making them accessible through a new interface for a traveller's relationship with a hotel.
Separating interface from operation also allows the same capability to be used through the website, Sarai or an external agent, according to the provider. Hotel content and commercial conditions remain behind those access points. For a brand offering different forms of customer service, the architecture seeks to maintain a consistent offer while changing the way customers ask questions, compare options and request an action. The channel's operational foundations are shared across those interactions.
Product information participates in conversion
Knowledge brings together information that can determine a purchase: room types and their differences, services, opening hours, policies, restrictions and common questions. In the provider's explanation, it also supplies context about dining, activities and loyalty benefits. Its commercial role extends beyond writing a hotel description. It supports answers to details arising during a decision, when travellers need to know whether an option suits the stay they are considering.
The infrastructure separately retains the engine's logic: inventory, prices, promotions and commercial rules. A room recommendation needs to move from description to a quote for the requested occupancy. Combining these elements connects what is communicated about a product with what can be booked. That relationship matters in a conversation where a customer may add preferences, change conditions or ask about the difference between two options without starting again at the beginning of the journey.
Sarai is the provider's own conversational interface over that system. It can show room photographs and promotional content alongside text, check availability and work with the hotel's rules. Mirai advertises responses in up to 100 languages and a multichannel presence encompassing website chat, WhatsApp, email, telephone and Instagram. These are capabilities of its offer described in March's earlier announcement, explaining how conversation fits into the broader direct-channel strategy and the different contact points a hotel can provide.
Conversational booking and card payment have distinct steps
Completing the operation deserves a more precise explanation than the idea of doing everything inside a chat. In Sarai's documentation, Mirai says payment uses BTP, Book, Think & Pay , the booking engine's environment where customers enter their card on a protected page. The described window is one hour or the period configured by the hotel. Conversation leads to an intention to book, while payment uses that particular process within the wider reservation journey.
The company says the process uses each property's configured payment processor. This helps explain the meaning of a conversational reservation: customers can express a requirement, receive a quote and progress through dialogue, while card handling retains its own page and payment environment. March's documentation explicitly distinguishes that moment from the exchange of messages. It does not present a card number typed into the conversation as the method of collecting payment from the traveller.
That step matters when understanding Mirai's proposed funnel. An interaction can begin with a question about services, continue with room selection and reach a confirmation of intent. The engine and payment environment complete the transactional dimension. The new interface changes how a stay is requested and managed, while the offer continues to need commercial conditions and an operation processed by the hotel's infrastructure. The conversation and the transaction have connected roles within that process.
Sarai’s documentation extends the relationship to other stages of a trip too. Before a stay it includes questions about opening hours, dining and services; during it, enquiries about places of interest, menus or getting to the airport; afterwards, matters involving invoices or complaints. These are examples of the service scope the provider describes for its own assistant. They connect purchase with the service surrounding a stay and with questions a returning customer may ask. Within that relationship, the knowledge base also supports answers beyond the specific moment of comparing a rate. September’s announcement adds ChatGPT access to that broader account of how a hotel’s capabilities can be used through conversation, from finding product information to managing a relationship around the stay.
The context for direct sales in online accommodation booking
Data on provider use helps explain competition for that guest relationship. A Statista Consumer Insights survey of 1,598 Spanish residents aged 18 to 64 who had booked accommodation online in the preceding twelve months found that 73% had used Booking.com and 30% Airbnb. A further 13% said they had booked directly with a hotel or private accommodation. The fieldwork covers July 2024 to June 2025 and provides context preceding the announcement.
Online accommodation providers used: three survey responses
Percentage of 1,598 Spanish residents aged 18–64 who booked accommodation online in the preceding twelve months. Three selected providers or provider types; July 2024–June 2025 survey period.
Tap or hover for details. All values are also available in the data table.
Collapse notes
Multiple responses: do not sum percentages. Reported respondent reach, not booking or revenue market share. The direct category includes hotels and private accommodation; the population does not represent all guests at Spanish hotels. Historical context, not Mirai results.
Explore the data table
| Category | Respondents who used it |
|---|---|
| Booking.com | 73% |
| Airbnb | 30% |
| Direct with hotel or private accommodation | 13% |
The survey allowed multiple responses: the same person could have used more than one option. Its percentages express reach among those respondents rather than revenue shares or a distribution of booking counts. The direct category includes hotels and private accommodation within an online-booking question. It also does not by itself represent every visitor to Spanish hotels, because the studied population is Spanish residents who purchased accommodation through the internet, rather than all demand arriving at properties in Spain.
The comparison frames the commercial issue that interests Mirai: maintaining a direct relationship when accommodation choices involve different providers and environments. The new ChatGPT access adds a route for using a hotel's own offer. The survey and deployment concern different moments and measurements. The data describes an earlier pattern of provider use; the announcement describes a new access and operational capability that the provider is beginning to deploy. Their relationship is contextual rather than a measurement of the new connection's performance.
Returning customers and people still looking
Mirai identifies a particularly defined application among returning customers. Someone already familiar with a chain could consult forthcoming stays, search for another property, apply loyalty benefits or manage a reservation through their assistant. The company compares that function with the role hotel-chain mobile applications have played in maintaining persistent direct access. An existing brand relationship gives context to the request and to the property the customer wants to consult, rather than beginning with the discovery of an unfamiliar business.
That scenario differs from discovering an unknown hotel. Mirai's announcement raises questions about how assistants will select connections and whether those connections will participate in inspiration and recommendation. The infrastructure can supply structured information and operations, but the role of an access point depends on how travellers find and use it. The company presents the initial deployment as a way to learn from those real behaviours and understand which capabilities people use within the conversational experience.
The distinction separates two marketing tasks. A brand needs travellers to consider it and also needs people who have chosen it to make a purchase or return. The announced access sits particularly close to the second task: consulting product, using conditions and making a reservation with the hotel. Potential participation in inspiration is a direction Mirai is exploring, rather than a result that follows automatically from having a connection available to an assistant.
Conversation supervision is part of the channel too
Lobby provides control of conversational operations within the architecture. Mirai's documentation describes conversation supervision, conversion and sales measurement, lead generation and management, and an opportunity for human intervention. A hotel can adjust the assistant's tone and behaviour and use information emerging in conversations to improve its knowledge base. Customer service and commercial activity are consequently connected with a management tool provided as part of the same system supporting the booking engine.
Human intervention has a defined role in that design. Staff can move from AI mode to human mode when necessary, while structured information supports automated answers. Knowledge about the hotel can be enriched through questions detected during activity. In business terms, the conversation becomes more than an exchange ending when an answer is delivered: it forms part of a system where a contact can be managed and its relationship with a sale followed. The provider places supervision alongside the ability to conduct the dialogue.
For this first phase, Mirai directs interested clients to their account manager. The announcement envisages gradual expansion to other properties, with limited deployments until relevant learning has been gathered. That commercial decision fits the project's scope: beginning to observe how people use direct-channel functions inside an assistant. The first hotels and chains give the provider a defined application setting for studying questions, operations and forms of interaction before extending that availability more widely to its customer base.
Who's who and how to read the figures
Mirai is the direct-channel technology provider making the announcement. Its legal notice identifies Mirai España SL, tax number B80039340, registered in Madrid's commercial register. ChatGPT is the conversational environment receiving the described connections. Hotel Emperador , Hotel Moderno , Porcel Hoteles and Gloria Thalasso & Hotels are the brands Mirai names in the first deployment; they do not constitute a count of every property belonging to those hotel chains or a statement that the capability has reached its whole customer base.
Knowledge, Sarai and Lobby have different functions: structured content, conversation and supervision. MCP is the connection protocol the provider uses to expose capabilities to external assistants. BTP is the payment environment described in Sarai's documentation. September's announcement rests on technical precedents published in March 2026 and December 2025; those dates remain separate from the new ChatGPT, from OpenAI , connection. The names describe different components involved in accessing information, managing interactions and processing a reservation.
The Booking.com, Airbnb and direct-booking percentages come from Statista's survey and measure provider use reported by its participants, with multiple answers permitted. They are not results for Mirai or the hotels in the deployment. The case's significance lies in the new mechanism: making content, availability and direct-sales operations accessible through a conversation while retaining the conditions and functions of the system managing the reservation behind that interaction. The announcement concerns that operational connection and its initial deployment.
In this article
Marian Vasile
Marian Vasile is a MarTech and growth professional. His work spans product marketing, SEO, paid advertising and deep online tracking, including behavioral analytics and attribution.