33% of Travel Sites Are Invisible to AI — Here's Why That's a $100B Problem

If you've built an AI travel agent — or used one — you've probably noticed something odd. Ask it to find the cheapest flight from London to Barcelona, and it'll show you British Airways at £180. Meanwhile, Ryanair has the same route for £29.

Your AI agent can't see Ryanair. Or Wizz Air. Or EasyJet. Or any of the dozens of low-cost carriers that carry 33% of all European passengers.

The Root Cause: No APIs

Traditional airlines distribute through Global Distribution Systems — Amadeus, Sabre, Travelport. These GDS platforms have APIs. Every travel app, every OTA, every AI agent connects to them.

But low-cost carriers opted out. Their business model is direct-to-consumer:

• Ryanair — #1 airline in Europe by passengers. No public API.
• Wizz Air — 80M+ passengers/year. No public API.
• EasyJet — 100M+ passengers/year. No public API.
• Transavia, Vueling, Norwegian, Volotea, Eurowings, Pegasus — same story.

That's not a minor gap. In Europe, LCCs carry one in three passengers. And they're usually the cheapest option.

What This Means for AI Agents

Every AI travel agent today — whether it's built on Claude, GPT-4, or a custom LLM — uses one of these approaches to search flights:

1. GDS API (Duffel, Amadeus) — misses all LCCs
2. Aggregator API (Kiwi.com, Skyscanner) — sometimes includes LCCs via scraping, unreliable
3. Screen scraping — breaks constantly, legally grey, can't actually book

None of these reliably cover LCCs. The result? AI agents show users flights that cost 3-6x more than what's actually available.

The Scale of the Problem

Let's put numbers on this:

• European air travel market: ~$100B/year
• LCC market share: 33%
• That's $33B in flights that AI agents can't access
• AI-powered travel booking is growing 40%+ YoY
• By 2028, AI agents will handle an estimated 25% of all travel bookings

If AI can't see a third of the market, that's billions in lost value — for travelers who pay more, for agents who lose users, and for LCCs who miss distribution.

The Fix: Human-in-the-Loop

At RunRelay, we took a different approach. Instead of trying to scrape airline websites (which breaks) or waiting for APIs that will never come, we built a human-in-the-loop execution layer.

Here's how it works:

AI agent sends a search request. RunRelay checks Tier 1 first (instant GDS results). Then checks Tier 2 (human operators search LCC websites in real-time). Returns combined results. The agent sees every flight, from every airline.

Why Not Just Wait for APIs?

LCCs have actively resisted API distribution for over a decade. Their model depends on ancillary revenue (bags, seats, priority boarding) that's hard to represent in GDS. Ryanair has sued screen scrapers. EasyJet shut down partner APIs.

The APIs are not coming. The gap will only grow as AI travel agents become the default way people book flights.

For AI Agent Builders

If you're building a travel AI agent and want complete airline coverage, RunRelay provides an MCP server and an OpenAI-compatible API:

One API call. All airlines. GDS + LCCs. Instant results for Tier 1, async callback for Tier 2.