How AI is replacing the results page at every stage of travel planning. Propellic watched US travelers plan trips with AI for a week and learned where brands should be competing. Full report and live webinar, September 30.
.png)
Propellic ran a week-long behavioral study following US travelers as they planned real trips end to end. We watched five stages: destination exploration, hotel research, multi-day itinerary building, day tours, and the handoff to booking.
Participants used their own AI tools, mostly ChatGPT, Gemini, Google AI Mode and Claude. They recorded their screens and talked aloud as they worked. All had booked leisure travel in the prior six months or were actively planning a trip.
ChatGPT took 47% of all tool use, more than Gemini and Google AI Mode combined. Most travelers stayed inside a single tool for the full week rather than opening a second AI to check the first. A traveler who does not find you in their tool does not go looking in another one.
AI pointed 71% of participants direct to the operator, and 68% independently preferred booking direct anyway. The framing and the preference reinforce each other, which makes this the most durable signal in the study. Our 2025 research found the same direction by a different method.
Why travelers hand AI the research, then take the transaction back at payment.
What travelers pointed to when asked why they trusted a recommendation.
When brand recognition starts working against you in an AI answer.
Why reviews and photos decide hotels, and your own copy barely registers.
How travelers read sponsored placements, and when an ad helped.
What happened when travelers were asked to build a multi-day itinerary.
Direction for hotels, operators, destinations and OTAs, with what to do about each finding.
Where this confirms our 2025 study, and where it changes it.