Expedia's image-selection automation is the concrete piece behind its AI marketing story

What Expedia actually put into production

Most of this conversation is a CMO discussing how the marketing function is changing. But two concrete deployments sit in the closing lines: Expedia Group uses OpenAI APIs to automate identifying and correctly presenting images to customers searching its travel sites, and it uses ChatGPT across its marketing department for creative use cases.

The image work is the most specific claim in the piece. Koedijk frames it in terms of scale — a catalog of hundreds of thousands of properties and vacation rentals — and quality control. He says the company uses OpenAI tools to moderate content quality so only high-quality, relevant assets get shown.

That is a narrow, checkable use case: matching and filtering images against a listing at catalog scale. It is not the same thing as "AI transforming marketing," and it's worth keeping the two apart when reading the rest.

The three-part breakdown, and where it's still aspiration

Koedijk splits AI's role into insights and analysis, productivity and creative development, and consumer behavior. The first two are described as current practice: summarization, trendspotting, real-time analysis, and producing text, image, and video content at scale, including brand ads.

The third — consumer behavior — is where he shifts to forecasting. He points to younger users moving from traditional search to tools like ChatGPT, and argues SEO alone is no longer enough for traffic acquisition. This is a stated concern, not a deployed solution.

Generative AI has given me a lot more scale.Montana Labs

That quote is the honest summary of what's shipped: scale in content and analysis. The agentic travel scenarios he describes — telling an agent "I need to go again" and having it handle a business trip — are explicitly framed as future possibility, backed by demos built with OpenAI and others, not live products.

Why Expedia keeps booking separate from inspiration

The most strategically interesting admission is where Koedijk draws the line. He says generative search is powerful at the inspiration stage — collapsing dozens of searches into one back-and-forth — but that when it comes to booking, people still gravitate to trusted platforms.

That's why he brings up One Key, Expedia's loyalty program with tiers that reward repeat usage. The logic is that AI can handle discovery, but the transaction and the customer relationship stay on Expedia's own platform. It's a defensive framing: use AI for the top of the funnel while protecting the part that monetizes.

He calls agents an "and, not an or" — brands building their own agents to complement human interaction. Read against the booking comment, that reads less like enthusiasm and more like a plan to avoid ceding the customer relationship to a third-party assistant.

The organizational move: pairing ML scientists with creatives

Koedijk describes a specific structure rather than a slogan. Expedia has a full-stack in-house marketing team that includes a marketing ML science group, and he says fluency came from building cross-functional teams — creatives paired with ML scientists, software developers, and marketers on focused use cases.

His argument for why this works is that generative AI lowers the barrier: you no longer need to be a data scientist to get value. That claim is consistent with the image-moderation deployment, which is an API integration a mixed team can own rather than a research project.

It starts with AI as a means to an end, and it's not an end goal by itself.Montana Labs

The takeaway: a content-supply problem solved before the demand-side bet

Expedia's shipped work addresses a supply problem — turning an enormous property catalog into correctly matched, quality-checked images and video faster than manual methods allowed. That's a real operational win with a clear metric to watch: engagement and credibility from showing the right assets.

The demand-side bet — that discovery is migrating into conversational tools and that SEO won't carry traffic — is still unresolved in this interview, hedged by the loyalty program and the insistence that booking stays on trusted platforms. For applied teams, the useful signal here is the sequencing: Expedia automated a bounded internal task first and is treating the consumer-behavior shift as an open experiment, not a solved one.