RESEARCH · OUT SEPT 30
The New Playbook for Travel Marketing in an AI World
We watched US travelers plan real trips with AI for a week. They chose destinations, researched hotels, built itineraries, found day tours, and tried to book, all on their own tools, screens recorded, thinking aloud.
One finding, ahead of the full report.
AI sent 71% of travelers direct to the operator. Independently, 68% said they preferred booking direct anyway. Our 2025 research found the same finding, but we since have expanded the study beyond just Google AI Mode. That means that travelers researching, planning and booking with LLMs are going direct, regardless of platform.
For operators that is the best distribution news in a decade, and it only pays out if there is something discoverable for AI to find. For everyone sitting between the traveler and the operator, it is a harder read.
There are seven more findings in the report, with what to do about each, and separate direction for hotels, tour operators, destinations and agencies.
The full study will be published on September 30th, with a live webinar the same day.
PAID MEDIA
You Cannot See What You Are Buying with ChatGPT Ads
by Melissa Spaulding - Associate Director of Paid Media @ Propellic
Somebody is going to ask you this quarter whether to move paid search budget into ChatGPT Ads. Before you answer, read this.
In Google Ads you see the search term, the keyword that matched it, and years of history telling you what that term is worth. In ChatGPT Ads you supply Context Hints, topics and descriptions of conversations that might be relevant, and OpenAI decides where your ad runs. You never see the conversation.
That closes the loop a paid search team actually works in. No search term report to mine, no negative list built from what you learned, no restructuring around a pattern you spotted.
For travel it costs more than most. "Things to do in Lisbon in March with kids" carries the destination, the season, the party composition and the planning stage in eight words. That density is what makes travel search buyable, and it is exactly what you give up.
The bigger error is treating it as the same job. ChatGPT sits higher in the funnel than paid search does. Someone comparing options in a conversation is not someone typing a booking query, and a channel that reaches them earlier shouldn't be measured against one that catches them at the end.
Worth knowing what the biggest spenders in travel are doing while everyone debates reallocation. OTA marketing spend rose again in Q2, at the same companies investing in AI visibility. Nobody with real data is moving money out of paid. They're funding both.
Our take: ChatGPT Ads need dedicated test budget. Not a slice of branded search, not a trim off the non-brand campaigns already producing qualified leads. A separate line, with its own hypothesis written down before you start, and a defined result that would justify keeping it. Without that, a test doesn't conclude, it just continues.
And go in knowing how little you'll see. Right now you're buying placement into conversations you can't read, on a platform with no performance history to lean on.
Sources: Osmundson, Search Engine Journal · PhocusWire
SEO & GEO
The Pages You Already Own Are Doing The Work
For a year the GEO advice has been the same. Be on Reddit. Be on YouTube. Show up where the community already trusts.
A study from Ten Speed points somewhere else. Across 170 prompts and 7,387 citation appearances, product pages took 24.1% of all citations, the largest single category. Reddit, YouTube, forums and discussion threads combined for 4.2%. Across the study, brand-controllable content, the pages a marketing team writes and owns, accounted for 88.3% of everything the AI models cited when buyers were deciding.
There are some important caveats from Greg Jarboe's outline. The dataset is B2B SaaS and professional services, and so doesn't translate directly to travel. It's also a single point-in-time pull rather than a tracked trend, and the study's author states that headline split was never statistically tested.
So although you should quote 24% at your next leadership team meeting, this information is still directionally helpful. From a travel leader perspective, Wyndham now feeds live data straight into ChatGPT, Claude and Gemini, then runs AI to watch what the models say back, and IHG is talking publicly about making its portfolio legible to them.
Here is a translation for other travel leaders - in travel terms:
Your product pages are tour pages, property pages, room types. Your comparison content is the three-day against the five-day, private against shared, this neighbourhood against that one, and most operators keep those in a sales deck instead of publishing them. Your directory profiles are your OTA listings and your Google Business Profile, which most brands treat as review inboxes rather than as pages that get quoted.
Our take: start with the twelve pages carrying the departures you actually want to fill. Not the whole site, not the blog. Citations on a departure you're retiring cost you money.
For each one, publish the comparison you currently email to prospects, and make the specifics machine-readable: duration, group size, what's included, what isn't, the price. Then check your OTA listing and your Google Business Profile say the same things as your own page. When they don't, the model picks one, and it might not be yours.
That's a quarter of work, not a replatform. And the structured data it produces is the same data feeding your paid feeds and deciding whether an agent can transact with you at all.
Sources: Jarboe, Search Engine Journal · Skift
Want To Level Up Your Travel Marketing?
Subscribe to the NavLog, our bi-weekly travel marketing roundup, where you’ll be the first to know about breaking news that impacts travel marketers and access exclusive performance marketing strategies and practical tips you can implement from the marketers at the leading edge of the travel industry.
