AI Travel Agents Tested for Vacation Planning Success

The latest wave of AI tools, including OpenAI's Operator and Anthropic's Computer Use, promise to streamline travel planning by handling everything from transport to accommodations. But how effective are they in practice?

The Experiment

To test these claims, the author tasked Operator—available to ChatGPT Pro subscribers—with planning a budget-friendly weekend trip with a focus on good food and art, preferring train travel. The AI suggested Bruges after ruling out Paris due to the author's recent visit.

• Train Booking: Operator successfully found Eurostar tickets to Brussels with onward travel to Bruges, though it initially selected inconvenient early-morning trains. After user intervention, it adjusted the schedule.
• Hotel Booking: The AI scoured Booking.com, eventually choosing Martin’s Brugge, a three-star hotel praised for its location.
• Itinerary: Operator faltered here, offering a generic one-day plan seemingly lifted from a vegetarian travel blog. Day 2 suggestions were vague ("visit any remaining attractions").

On the Ground

During the trip, Operator struggled with real-time tasks like finding the correct train platform in Brussels. The author resorted to station displays while the AI lagged. For a better itinerary, the author turned to ChatGPT, Google’s Gemini, and Anthropic’s Claude, which provided more detailed hourly plans—though these still felt like standard tourist routes.

• Dinner Reservation: Operator excelled here, booking a table at an authentic, off-the-beaten-path restaurant themed around pigeons—a gem not found on mainstream platforms like TripAdvisor.

Key Takeaways

• Pros: AI agents can reduce decision fatigue by curating options (e.g., hotels, trains) and uncovering hidden gems.
• Cons: They lack common sense (e.g., early trains), struggle with multi-step tasks, and often default to generic itineraries.
• Human Touch: As Emma Brennan of ABTA notes, travelers still value human agents for reliability when things go wrong.