One Internet. Two distribution ecosystems

For the last three decades, hotel distribution has evolved around a single digital ecosystem. Travelers searched, compared alternatives, visited hotel websites or OTAs and eventually completed a reservation. As an industry, we have spent years optimizing every stage of that journey through conversion funnels, attribution models, SEM, SEO, metasearch and especially direct booking strategies.

At Mirai we don't believe that ecosystem is disappearing. At least not today. However, we do believe a second distribution ecosystem is beginning to emerge alongside it. In this new environment, travelers will increasingly rely on AI assistants to search, compare and even complete bookings on their behalf. The result is not the end of the web, but the emergence of a second distribution ecosystem built around very different principles.

Whether these ultimately become two distinct ecosystems or two interfaces over the same digital infrastructure is almost secondary. What matters is that hotels will increasingly need to compete under two fundamentally different interaction models.

What changes?

• Human → Agent through delegation: in the web ecosystem, travelers search, compare and book themselves. In the agent ecosystem, they increasingly delegate these tasks to AI assistants, who act on their behalf while keeping the traveler in control of the final decision.

• Websites and web services→ Model, Knowledge and Capabilities: the web is built around documents designed to be read by humans. The agent ecosystem increasingly relies on structured knowledge and executable capabilities that machines can understand, trust and use directly.

• Search → AI Assistants: traditional search engines are complemented by conversational interfaces that understand intent instead of matching keywords.

• Keywords → Prompts: optimization shifts from isolated search terms to natural language requests and contextual conversations.

• Preview and Blue Links → Answers and Citations: instead of lists of links, assistants generate direct answers while citing trusted sources when appropriate.

• Limited and Session-based → Contextual and Conversational: traditional websites personalize experiences using relatively limited signals such as location, device or browsing history. AI assistants can adapt every interaction using the traveler's intent, preferences, past conversations and real-time context.

• Organic and Paid Placements: visibility will likely continue to combine organic relevance and sponsored placements. The monetization model may evolve, but advertising is unlikely to disappear.

• Clicks → Conversations and Executions: success is measured less by traffic and more by completed conversations and successful task execution.

• Documents → Knowledge Networks: websites organize information into pages designed for humans. AI agents increasingly require structured, contextual and dynamic knowledge that machines can consume directly. The critical challenge for hotels won't just be creating this knowledge, but managing its distribution -either through their own integrated tech partners or open protocols like MCP- to prevent major OTAs from once again monopolizing the source of truth.

• SEO → Agent Optimization: ranking in search results is complemented by becoming a trusted, authoritative and executable source for AI agents.

• Web Transactions → Agentic Transactions: reservations evolve from users completing forms to AI agents executing transactions on the traveler's behalf.

• Browser: the browser may no longer be the primary gateway to hotel distribution as AI interactions expand across apps, operating systems and other interfaces. 

• Tracking: web analytics are built around sessions, clicks and page views. The agent ecosystem will require entirely new metrics to measure visibility, conversations, tool calls, executions and completed transactions.

The purpose of this comparison is not to suggest that every one of these changes will happen overnight, nor that the ecosystem on the left will disappear. Rather, it illustrates that hotel distribution is expanding into a second environment where many of the concepts and metrics we have relied on for years will gradually be complemented by new ones. We already understand how to compete in the web ecosystem because we have spent decades refining it.

The next question is therefore not whether AI will replace the web. It's what hotels will need to build in order to compete in this second distribution ecosystem. Travelers will decide which one to use. 

Is the web really the foundation of the agent ecosystem?

One of the most common assumptions today is that AI assistants will simply become better at browsing hotel websites. If that's true, then the obvious strategy is to optimize existing websites for LLMs through GEO, AI SEO, Web MCP or similar approaches.

We believe that's only part of the story, and probably a transitional stage rather than the final architecture.

The web was designed to publish documents that humans can read and navigate. AI agents operate differently. They need information that is complete, contextual, continuously updated and reliable enough not only to answer questions, but eventually to execute actions on behalf of travelers.

After analyzing millions of conversations handled by Sarai,  our native AI Assistant, we have found that only around 30% of guest questions can typically be answered directly from a hotel's website. The remaining 70% depends on operational knowledge that rarely exists on public web pages: hotel policies, operational procedures, exceptions, services and countless practical details that travelers need before, during and after their stay.

This isn't a criticism of hotel websites. They were never designed to become operational knowledge bases. Their role is to inspire, market and convert visitors, not to expose every operational detail of a property.

The same applies to traditional Content Management Systems. CMS platforms were built to publish web pages, not to expose structured operational knowledge or live transactional data that machines can consume directly. They will undoubtedly evolve, but we don’t believe the web server itself will become the primary infrastructure through which AI agents interact with hotels.

That doesn't mean hotels will have to build this new infrastructure themselves. Just as hotels rely today on specialized providers for their booking engine, CRS or PMS, those same technology partners will increasingly become the bridge between hotels and AI assistants. Their role will be to expose trusted knowledge and transactional capabilities in whatever standards and protocols the market ultimately adopts.

This is already the direction we have chosen at Mirai. Every hotel connected to our platform already has its own MCP endpoint, and we have also built a provider-level MCP that represents the direct channel of thousands of hotels—similar to the role we already play today as a Google Hotel Partner. Whether MCP becomes the dominant standard or eventually evolves into something else is almost secondary. Our objective is much simpler: ensuring that every connected hotel is ready to participate in the emerging agent ecosystem without having to rebuild its technology stack. In our view, this should become as standard as exposing rates to Google Hotel Ads or connecting to a metasearch platform.

The challenge for hotels, therefore, is not to become AI companies. It's to ensure that the technology partners they choose are preparing them to compete in this emerging ecosystem.

How do hotels optimize for the emerging AI ecosystem?

Over the last 18 months, the AI landscape has evolved at an extraordinary pace. New models, protocols and standards have appeared almost every week, making it difficult to distinguish temporary experimentation from long-term direction. Fortunately, the picture is starting to become clearer.

In a recent article, "How a hotel appears in AI assistants: the three layers of visibility," we  explain how hotels become visible inside today's AI assistants. If you haven't read it yet, we highly recommend it.

In short, visibility is no longer determined by SEO alone. It increasingly depends on a combination of the model's own knowledge, web search and grounding through trusted external sources. That framework explains how hotels become visible. This article focuses on what comes next.

As AI evolves from answering questions to acting on behalf of travelers, hotels will need more than visibility. They will need an infrastructure that allows AI agents not only to discover them, but also to trust them, understand them and interact with them.

From visibility to interoperability

The three visibility layers explain how AI assistants discover a hotel. The next challenge is interoperability.

Our hypothesis is that every hotel will eventually need four fundamental building blocks to compete in this emerging ecosystem. The technologies implementing them will undoubtedly evolve, but I believe these underlying requirements will remain remarkably stable.

1. Identity

Before an AI agent can trust a hotel, it must first know that it is interacting with the hotel's official digital source.

Just as domains became the canonical identity of organizations on the web, the agent ecosystem will require mechanisms that allow AI assistants to verify that a hotel's knowledge and capabilities are being exposed either directly by the hotel or through an authorized technology partner.

Identity is therefore not about content or transactions. It's about proving who is the legitimate source of truth.

2. Authority

Being the official source is not enough. AI assistants must also determine whether that source deserves to be trusted above others. Authority will depend on factors such as reputation, data quality, freshness, breadth of exposed capabilities, successful executions and, potentially, the commercial or contractual relationship between the hotel and the technology provider.

Exactly how authority will be established remains uncertain. However, much like search engines developed ranking signals over time such as Google’s PageRank, AI assistants will inevitably develop mechanisms to decide which providers deserve to act on behalf of users.

3. Knowledge

AI agents don't need more web pages. They need reliable knowledge.

That knowledge extends far beyond marketing content. It includes operational policies, procedures, services and the countless details that make every hotel unique. More importantly, it must be structured, continuously maintained and exposed in formats that machines can consume directly.

4. Capabilities

Finally, AI agents won't simply answer questions, they will execute tasks.

Checking real-time availability and pricing, making or modifying reservations, identifying loyalty members, downloading an invoice, processing payments or handling cancellations are not pieces of content. They are capabilities that hotels will need to expose securely through their technology partners so AI agents can interact directly with hotel systems.

Taken together, these four building blocks describe what I believe an agent-ready hotel will look like. They don't replace websites or existing digital channels. They complement them with something fundamentally new: a digital presence designed not only to be discovered, but also to be trusted, understood and used by AI agents.

Conclusion

The web isn't going away. For years to come, it will remain the primary distribution ecosystem for human travelers, and SEO will continue to be one of the best investments hotels can make. 

At the same time, AI assistants are rapidly becoming the preferred interface for inspiration and discovery. The booking funnel, however, is still incomplete, which means the web continues to dominate consideration and transactions. However, this is about to change.

And that change may be arriving faster than expected. In August 2026, Google introduced hotel booking directly within AI Mode in Search, describing the experience as a way to “discover and book your next hotel in one seamless conversation.”

This is an important milestone. For the first time, one of the world's largest travel discovery platforms is explicitly connecting conversational discovery with hotel transactions inside an AI interface. The agentic booking funnel is no longer just a theoretical architecture. It is beginning to take shape in front of us.

The real transformation, however, will happen if AI assistants evolve toward organic discovery rather than relying exclusively on commercial intermediaries. At Mirai, we believe they will, and the first signals are already visible. Claude is experimenting with MCP submissions, OpenAI has introduced Apps, and Google is working on Agentic Resource Discovery (ARD). These initiatives are different in their implementation, but they all point in the same direction: an ecosystem where AI agents can discover, evaluate and interact with reliable providers beyond traditional web navigation.

If that ecosystem matures -and I believe it will- the industry won't simply need better websites. It will need technology partners capable of making hotels agent-ready.

Hotels won't need to become AI companies. Just as they don't build their own booking engines, CRS or PMS, they won't build their own AI infrastructure. Their competitive advantage will increasingly depend on choosing partners capable of exposing their knowledge and capabilities to whatever agent ecosystem ultimately emerges.

The web gave hotels digital visibility. The emerging agent ecosystem will require digital interoperability.

Hotels won't choose between these two distribution ecosystems. Travelers will.

The winners won't be the hotels with the best websites. They will be the hotels whose technology partners have enabled them to compete in both ecosystems.