Somewhere in the fine print of Expedia Group B2B’s Explore ’26 announcement sat a detail that mattered more than the toolkit headline. Expedia opened its inventory through an MCP server, the connector standard that lets AI agents like ChatGPT, Claude, and Perplexity query a platform’s data directly. The travel press treated it as a footnote to the AI toolkit story. It should not have been.
The Question Nobody Finished Asking
The airline industry has spent a decade debating who controls the offer: the airline, the GDS, or the agent. NDC was built to settle that question in the airline’s favor, or at least to give carriers a mechanism to try. What nobody fully priced in was a fourth party asking for the offer: an AI agent, acting on a traveler’s behalf, that does not care which distribution channel produced the answer, only that it got one fast and correctly formatted.
Expedia’s MCP move is a bet on where that traffic goes, not a claim that it has already arrived. If a traveler asks a general-purpose AI assistant to find and book a flight, the assistant needs a source it can query programmatically. Expedia is positioning itself early to be that source, machine-readable before the volume exists to justify the investment. When it works, the airline, the GDS, and the NDC aggregator sitting behind that fare become invisible to the transaction. The agent sees Expedia. It does not see the plumbing.
Why This Is Not the Same Story as the AI Toolkit
TDN covered the structural repositioning in the toolkit and CarTrawler announcement in May. That piece argued Expedia was building the developer and partner layer, not fighting the GDS for the agency relationship. The MCP server is a narrower and, in some ways, more consequential move. It is not about partners embedding Expedia into their own products. It is about Expedia becoming legible to AI agents that are not Expedia’s partners at all, that a traveler might reach through any interface, with no commercial relationship between the two companies.
That is a distribution channel airlines and GDS operators have almost no visibility into, let alone control over.
What This Means for NDC’s Underlying Argument
NDC’s case has always rested on the idea that direct, rich data control lets airlines differentiate their offer and capture more value. That argument assumes the entity making the booking decision cares about differentiation: seat maps, loyalty status, ancillary bundles, brand.
An AI agent optimizing for the user’s stated preferences, price and schedule fit primarily, may not reward that differentiation the way a human shopping on an airline’s own site does. If agentic search flattens the shopping experience into a small set of ranked options pulled from whichever source answers fastest and cleanest, the carriers with the most sophisticated NDC retailing capability do not automatically win the agent’s attention. The carriers with the cleanest, most queryable data feed do.
Expedia, sitting on decades of structured inventory and now offering it through the protocol AI agents are standardizing around, has a real structural advantage in that specific race. Airlines running NDC through their own direct channels do not have that advantage unless they also make their data available in agent-legible form, which is a build most have not yet started.
The Africa Angle, Again
The Africa Distribution Readiness Index work TDN has been building all points to the same gap: NDC adoption across the continent remains uneven, and few carriers or agencies have direct API infrastructure mature enough to serve traditional online booking flows well, let alone agentic ones. An MCP-queryable layer is a further step beyond where most African aviation distribution currently sits.
That does not make Expedia’s move irrelevant to the region. It makes it more relevant. Many African carriers and agencies remain focused on establishing the API and NDC foundations that agentic distribution would eventually require. If agentic AI becomes a meaningful booking channel over the next few years, that foundational work becomes the prerequisite for a layer that has not been built yet, not a substitute for it. The gap is not closing. It may be widening one layer up.
The Real Story
Whether AI agents actually reroute meaningful booking volume away from direct and GDS channels in the near term is still unproven. Expedia is positioning early, not proving a result.
But the positioning itself reveals something about where the industry’s long argument over distribution control is heading next. The GDS was built to control the workflow. NDC was built to let airlines reclaim control of the offer from the GDS. APIs extended that control to any partner willing to integrate directly. Each layer answered the question of who owns the previous one.
Agentic commerce introduces a new layer on top of all of it: the interface through which the shopping decision actually gets made. NDC was designed around the assumption that the airline controls distribution by controlling the offer. It said little about who controls the interface the traveler, or their agent, uses to see that offer in the first place. Expedia’s MCP server is an early answer to a question most airlines have not finished asking.
That is the uncomfortable one for the industry to sit with. Winning the NDC argument does not automatically win the next layer. Someone will own the interface through which agents shop. Right now, Expedia is one of the few companies actively building for that role.
