Amadeus Defends GDS Model as AI Booking Middlemen Multiply

Amadeus booked €5.08 billion ($5.67 billion) in group revenue for the first nine months of 2026, up 5.2% at constant currency, even as bookings fell 2.8%. The gap between those two numbers is the story: pricing, not volume, is carrying the distribution business, and the AI-agent question is now testing whether Amadeus keeps its seat at the table.

On the Q3 earnings call Friday, November 6, an analyst put the awkward question directly. Why is Meta's Muse — the AI assistant actually moving bookings — using Duffel, a flight-API startup, rather than Amadeus to find flights, complete bookings and handle changes?

CFO Carol Borg's answer was tidy. The Muse-Duffel partnership, she said, supports Amadeus' core belief: AI platforms will plug into aggregators rather than build booking systems themselves. "It actually reinforces it," she said of the GDS (global distribution system, the intermediary network connecting airlines to travel sellers) model.

One detail undermines the cheer. Duffel, not Amadeus, holds the direct platform relationship with Muse. Duffel's inventory stack includes direct airline connections, NDC content (IATA's New Distribution Capability, the airline-owned standard for direct retail offers) and traditional GDS supply from Travelport. In other words, Amadeus content may still flow through the pipe — but somebody else owns the pipe's front door.

Who else is crowding the middle?

The field of intermediaries positioning between AI assistants, banks, super-apps and suppliers is getting crowded. Some compete with the GDS; others quietly depend on its inventory.

• Selfbook powers hotel booking and payments for Perplexity.
• Nuitée supplies hotel booking infrastructure to Grab and Revolut and is building APIs aimed at AI applications.
• Katanox has partnered with Selfbook on distribution and settlement.
• AirGateway and Kyte offer API-based airline connectivity.

CEO Luis Maroto's counterargument is scale. He pointed to roughly 3 billion daily travel searches and argued that servicing complex itineraries at that volume is harder than connecting to individual airlines one by one. A fair point, as far as it goes. It doesn't explain why Muse picked someone else.

What do the numbers say?

The financials show a business holding price while volume drifts. Revenue per booking rose 5% in the first nine months, against the 2.8% bookings decline, which management attributed largely to Middle East disruption. Adjusted EBIT climbed 5.5%.

Guidance is the tell. Management expects full-year results near the lower end of its range, while maintaining the 2026–2028 targets. That's a company confident in its margins, less so in its volumes — and volumes are where AI-agent disruption would bite first.

The structural question is simple. Amadeus has decades of airline relationships and unmatched scale, but scale and relationships don't automatically translate into a direct role when AI platforms choose their plumbing. Borg is right that someone must do the booking work. Whether that someone is Amadeus, or an API layer sitting between the AI platform and Amadeus, is the difference between owning distribution and supplying it.

What happens next?

Watch two signals in the coming quarters: whether Amadeus lands a direct integration with a major AI platform of Muse's stature, and whether revenue per booking keeps outrunning volume as agent-mediated shopping grows. The company's 2026–2028 targets assume it stays indispensable. Duffel's front-door position with Muse suggests the market hasn't fully agreed yet.