AI Travel Planners Don't Get How You Really Plan Trips

If you've ever asked an AI chatbot to plan a vacation, you may have noticed something odd: the trip it hands back feels generic, even when you answered all its questions. A new study from researchers at UC San Diego explains why. They interviewed eight everyday travelers and nine professional travel agents, then mapped out how real planning actually unfolds. The short version: almost nobody starts with a clear, stable wish list.

Instead, planning is a back-and-forth. You say "somewhere warm, under $2,000." Then you see flights are cheap to Portugal, so suddenly Portugal is the destination. Your budget shifts. Your dates shift. A friend mentions a restaurant, and that becomes a must-do. The researchers identified 11 distinct actions people take around these shifting limits and wishes — surfacing them, testing them, refining them, and juggling them against each other. Professionals do this instinctively; they probe, suggest, and adjust rather than gathering requirements once and disappearing.

Most AI travel tools do the opposite. They treat your first message like a final order form: collect the constraints, generate the itinerary, done. That works fine for simple requests, but it breaks down the moment your preferences evolve — which is basically always. The paper argues this mismatch is a big reason AI itineraries often feel like a list of tourist traps rather than a trip you'd actually enjoy.

The researchers offer design guidance for building better AI planning assistants — tools that ask questions at the right moment, notice when you've changed your mind, and let you and the AI adjust the plan together. Practical takeaway: the next generation of travel AI won't just be smarter, it'll be chattier, more patient, and better at changing course mid-plan. For now, treat AI itineraries as a first draft and expect to push back a few times.