AI, NDC and MCP: how air ticket distribution is changing
At the recently held conference Airline Distribution 2026 held in March in Barcelona There are open questions that concern the whole Aviation ecosystem Will artificial intelligence change distribution structurally, or will it just “coat” existing systems with a layer of optimisation, and can a standard like NDC become less important in a world of AI agents that understand complexity?.
The IATA NDC standard, first introduced in 2012, was also the focus of the discussion, developing at varying speeds over the years depending on the carrier, partners, and market.
Although it is a concept already well known to everyone, and the technology has been applied for a long time, it's worth reminding: According to IATA, NDC is a data exchange format (XML) based on the Offer & Order processes, enabling airlines to create and distribute relevant offers regardless of the sales channel.
Direct vs. indirect: the mix that doesn't disappear
Airlines have been pushing direct sales for years, but the market and passenger behaviour still “pull” towards Multichannel world. The discussion emphasised that, even if a carrier wished to sell all capacity exclusively through its own channels, the reality of demand, destination, and purchasing habits does not support this.
The concrete figures clearly illustrate why there is no universal "ideal" ratio. American Airlines has stated that its direct share is around 55 per cent and that this ratio varies depending on market conditions, while United Airlines today reports 70 per cent direct sales, with the remainder through NDC (outside the GDS) and legacy channels.
The key message for intermediaries (TMC, OTAs, and others) is simple: the indirect channel remains critical, but increasingly less as a “place where the price is shown,” and more as a layer that manages content complexity, rules, and servicing. It is precisely servicing that is becoming the key differentiator today, as confirmed by the agents I speak with in the region.
NDC as a transition towards Offer & Order to the world and impact on EDIFACT
From many colleagues and partners, particularly at the last ITB, I hear about the gradual reduction in the importance of EDIFACT, which emphasizes the even greater impact of NDC and is in fact nothing new.
For an understanding of the problem, it is useful to know that modern platforms often carry “traditional” EDIFACT content and newer NDC content in parallel, with attempts to display everything in a unified shopping and booking experience. This is increasingly becoming the standard with GDSs and other distributors.
Although the direction is quite clear, the question remains as to how quickly the process will unfold. Differentiation in distribution methods by business type is also to be expected.
Certainly, the expectations are that the hybrid model will remain active for quite some time.
MCP: “agent-ready” web and the idea of universal adapters
How does MCP fit into this story? According to MCP's technical explanations, it is an open standard that allows large language models and AI agents to communicate with external tools and data sources (e.g. APIs, databases, services) in a structured manner.
WebMCP moves a step further towards the web: it emerges as a proposed standard for exposing “structured actions” on web pages so agents can reliably perform tasks.
This is where we arrive at the provocative thesis that is increasingly being mentioned: if an AI agent can “read” and understand the complexity of different implementations, how perfect does standardisation even need to be? It is often mentioned that integrations which previously took months can sometimes be done very quickly when a large part of the understanding and mapping is transferred to AI models.
Accordingly, MCP and WebMCP—as relatively new products/concepts—could eventually displace NDC. What is particularly interesting for distribution is that MCP is already taking a very “concrete” shape in air travel retailing. TPConnects Technologies, we wrote about them recently, integrated the MCP layer into its Astra NDC platform with the aim of acting as a universal adapter for different versions and variations of NDC schemas and to facilitate connectivity to AI agents.
Look-to-book: a cost that AI can both increase and decrease
When we speak about NDC and artificial intelligence, it is inevitable to also raise the question of distribution costs – especially through the look-to-book (L2B) ratio.
Look-to-book, or the ratio between searches and actual purchases, is becoming one of the key challenges. The higher the ratio, the higher the system and operational costs.
NDC has significantly increased this ratio from several hundred to thousands, while AI further intensifies the pressure through a higher volume of searches and automated testing. The problem is not solely technological. The question of business models is also increasingly being raised – who bears the cost of high query volumes.
The industry is already seeking solutions. Amadeus is using AI filters to reduce unproductive traffic, while Sabre and Travelport are attempting to reduce the number of unnecessary queries and alleviate system load through caching models.
What does this mean for users and travel tech in 2026?.
For agencies, TMCs and OTAs, 2026 brings a clear change: distribution and technology can no longer be viewed separately. The market is already hybrid today – part of sales goes directly, part through NDC, and part through legacy systems. Therefore, both sales and service processes must be adapted to this reality, not to “ideal” models.
At the same time, agentic AI is changing the way users book travel. Increasingly, solutions are moving towards conversational interfaces with integrated booking and payment, which could become the standard for major players as early as 2026.
Distribution in the aviation industry has become very complex, and no matter how much it seemed some time ago that the job of booking agents was becoming easier and more automated, quality service and passenger assistance require more and more time and effort. In fact, it all comes down to what each person wants and how much they are willing to pay 😊
