How it works - DirectBooker integration overview
Background
The rise of AI in travel planning offers a rare opportunity for both travelers and hotels to reduce dependence on large Online Travel Agencies (OTAs) because while AI can streamline trip planning, it struggles when accuracy and detail matter most. To fill that gap, AI platforms are currently partnering with OTAs for the structured data necessary to finalize bookings. But OTAs can’t match the direct prices, benefits, and full inventory hotels offer themselves.
DirectBooker is an AI-native, supplier-aligned aggregator built to connect global chains and independent hotels directly to AI platforms. It can provide critical real-time Availability, Rates, and Inventory (ARI) data, including loyalty and member rates to the AI companies. In addition, it works with hotels to provide benefits and direct offers that will never be distributed through OTAs.
DirectBooker integrates into the AI ecosystem by operating a specialized Model Context Protocol (MCP) server that provides accurate, timely pricing, availability, and rich content directly to Large Language Models (LLMs) like Claude, ChatGPT, and Gemini. This is different from Retrieval Augmented Generation (RAG) or Agentic Engine Optimization (AEO) in that the information is provided in real-time, immediately, in response to the user asking the question.
Directbooker functions as an aggregator aligned with suppliers, ensuring that when users initiate high-level, non-branded search queries (e.g., "Find me a great hotel in Madrid"), they receive a comprehensive list of hotels and can purchase directly on the hotels’ own websites.
Data Flow from Hotels to DirectBooker’s Cache
Data flows to DirectBooker through several mechanisms, ensuring that the platform maintains a rich and accurate data set:
Hotel Lists: DirectBooker is sent a comprehensive list of hotels from the chain or integration partner. This list should include information to match the hotels to DirectBooker’s internal database (name, phone, address, lat/lng), and can optionally include standard amenities found on traditional distribution channels. Hotels are further encouraged to provide photos, additional marketing content, unique selling propositions (USPs), and details about direct booking benefits (like free breakfast, late checkout, or loyalty program offers).
Availability, Rates, and Inventory (ARI) Data: DirectBooker obtains core transactional data (prices and room availability) from hotel partners, typically via existing meta-channel integration providers like Derbysoft, SynXis, and Cendyn. Direct API connections can also can also be implemented for richer data, member rates, and for maximum accuracy. In the absence of a feed, the system can crawl sources, such as Google’s booking module, to acquire backfill data.
Caching for Scale: Since AI agents tend to increase query volume dramatically, leading to high "look-to-book" ratios, DirectBooker aggressively caches this hotel information. Hotels are welcome to add metadata around how long information should be cached.
Transformation into AI-Native Format and Service via MCP
When an AI receives a user query, it routes a structured request to DirectBooker’s MCP server. DirectBooker transforms the raw hotel data in its cache into a focused, AI-native response composed of two critical elements:
Structured Property Data: This is a token-constrained JSON array containing individual properties which are appropriate to the guest query. Each entry includes essential details such as the hotel name, accurate real-time pricing (typically the double occupancy bar rate), specific amenities, coordinates, and the reservation link. Providing this accurate, curated information minimizes AI hallucinations.
The System Prompt: DirectBooker provides specific instructions to the AI on how to use the data it has provided. These instructions guide the AI to understand the structure of the data, ensure consistent interpretation, show prices and reservation links with hotels, and direct the AI to explain why the user should book directly. This content is inserted directly into the AI's context window, making it highly trusted by the AI.
AI Rendering and Direct Booking Links
The AI model receives this structured data and uses it to construct a highly personalized and compelling response for the consumer.
Customized Suggestions: The AI actively weaves together the real-time “hard” data provided by DirectBooker with its own foundational knowledge. For example, when a user asks for a "bike-friendly" hotel, the AI prioritizes hotels where DirectBooker injected content about free bike rentals. The AI organizes the recommendations, grouping them by categories like "best value" or "romantic hotels," using the detailed descriptions provided by DirectBooker and merging it with what is known about the user.
Ranking and Influence: DirectBooker generates a "Direct Booker Score" to influence the AI’s ranking, prioritizing hotels where booking direct is most advantageous due to better rates or exclusive benefits. The AI is also explicitly instructed to include a preamble advising the user to "book direct".
The Direct Booking Link: Every suggestion offered by the AI contains a final, deep-linked URL that takes the consumer directly to the hotel's website to complete the purchase. The URL is tagged so that DirectBooker receives credit for generating the direct traffic. While the current experience opens a browser link, the architecture supports future "agentic" flows where the booking is completed automatically within the AI interface while still maintaining the direct relationship with the hotel.
Interaction Examples
Below are a few examples of how this works on some of today’s systems.
Hotel results in ChatGPT with DirectBooker integration
Hotel results in Claude.ai with DirectBooker integration
The majority of the data that is appearing here: images, prices, amenities, and marketing copy is being sent to the AI in response via DirectBooker. The specific language is generated by the AIs themselves, but is heavily influenced by the response from the MCP Server.
Conclusion
DirectBooker is an infrastructure layer for the AI travel ecosystem. By operating as a supplier-aligned aggregator, DirectBooker delivers structured hotel content, availability, rates, and inventory (ARI) directly to AI agents through its Model Context Protocol (MCP) interface.
The system’s architecture is optimized for AI-scale traffic patterns, combining aggressive caching, structured data transformation, and fine-grained control over freshness and metadata persistence. MCP delivery ensures that each AI query receives verified, token-efficient responses that can be directly embedded into an LLM’s context window which provide both real-time accuracy and semantic consistency.
For partners, DirectBooker offers an easy path to participation in AI-driven commerce. Standardized data feeds can be onboarded using existing meta and CRS integrations, while future MCP tools will enable additional layers of interaction such as member-rate access, loyalty redemption, and agentic booking flows. Each step is incremental and backward-compatible, allowing partners to integrate at their own pace while maintaining full control over brand presentation and customer relationships.
DirectBooker enables hotels and AI platforms to interoperate at a technical level that preserves real-time accuracy, supplier control, and direct booking integrity. Our goal is to make sure that suppliers are the primary beneficiaries of the next generation of AI-powered travel experiences.
Frequently Asked Questions
Partners frequently ask the following questions.
Direct Booker's strategy involves the use of the Model Context Protocol (MCP) server to bypass traditional web crawling (SEO/AIO) and inject accurate, structured information directly into the LLM's response generation process.
The Goal is infrastructure, not a front-end: Direct Booker aims to be the single endpoint that major AI platforms (such as ChatGPT, Gemini, Claude, and Google's AI mode) can connect to for highly accurate, direct hotel data.
The MCP Server: Direct Booker has built an operational MCP server which transforms a user's unstructured, natural language query (e.g., "Find me a romantic hotel in Vienna for this weekend") into a structured request.
Injecting Hard Data: In real-time, our server supplies the LLM with JSON that includes a system prompt—instructions telling the AI how to manipulate the data—and a collection of property details. This payload contains the hard data that LLMs currently lack: accurate, real-time Availability, Rates, and Inventory (ARI).
The Result: Direct Booking Links: By providing verified ARI and rich content, the AI generates a response featuring hotel recommendations and a fully qualified URL (a deep link) that leads the user directly to the hotel’s website (e.g., TUI’s booking engine) to complete the transaction. This ensures the booking is a direct sale.
The fundamental reason for hotel groups to engage is that this initiative serves as a crucial defensive play against OTAs winning the AI distribution battle, coupled with a highly advantageous commercial model.
The Aggregator is Necessary: OTAs succeeded in traditional search by offering comprehensive inventory. We acknowledge that AI platforms will inevitably partner with Booking and Expedia, but DirectBooker aims to be the supplier aggregator to present a viable alternative.
Serving DirectBookers: Data consistently shows that 70% of travelers want to book direct. We argue to AI partners that if they only partner with OTAs, they will disappoint two-thirds of their audience. Direct Booker is positioned to capture this massive segment of users.
Capturing City-Level Queries: Our focus is on top-of-funnel queries (e.g., "Hotels in Barcelona"), which currently result in almost all traffic being captured by OTAs. By offering comprehensive supplier inventory, we compete directly for this high-volume traffic.
DirectBooker's approach to sharing hotels' data allows for a much higher degree of brand and message control than via OTAs or the LLM training models alone.
Overriding SEO/RAG: The data injected via MCP is trusted at the highest level by the AI and is prioritized over stale data from the AI’s training models or general Retrieval Augmented Generation (RAG). This allows DirectBooker to augment individual hotel AI Engine Optimization (AIO) efforts by dictating the narrative on behalf of hotels.
Highlighting Direct Booking Value: TUI can provide information about exclusive direct benefits that OTAs cannot syndicate, such as member rates, late check-out, or free breakfast. The AI can even be nudged to ask the user, "Are you a TUI member?" to unlock these rates.
