I Made AI My Travel Agent. I Wound Up in a Quaint Seaside Town.

Key Concepts

• AI-Driven Travel Planning: Using Large Language Models (LLMs) like Gemini to curate travel itineraries.
• Price-to-Beauty Ratio: A metric used to evaluate travel destinations based on affordability versus aesthetic appeal.
• Real-time Data Limitations: The tendency for AI to provide outdated or inaccurate information regarding logistics (transportation, business hours).
• Human-in-the-loop (HITL): The necessity of human oversight and verification when utilizing AI-generated plans.

1. The Experiment: AI-Planned 48-Hour Getaway

The objective of this experiment was to evaluate the efficacy of AI in travel planning by strictly adhering to an itinerary generated by Google’s Gemini. The user requested a "seaside getaway from London," resulting in the selection of Saltburn-by-the-Sea, a destination chosen for its high price-to-beauty ratio despite its relative obscurity among local travelers.

2. Logistical Challenges and AI Limitations

The experiment highlighted significant gaps in AI’s ability to handle real-time, physical-world logistics:

• Navigation Errors: The AI provided incorrect directions at the train station, nearly causing the traveler to miss their connection.
• Contextual Blindness: Upon arrival, the AI suggested a walking route to the hotel that ignored adverse weather conditions (wind and rain) and safety hazards (speeding traffic).
• Static Data Issues: The AI recommended a restaurant that was permanently or temporarily closed, forcing the traveler to rely on a "Plan B."
• Oversight of Local Attractions: The AI failed to suggest high-value local experiences, such as the mobile sauna and North Sea swimming, which the AI later acknowledged as a "massive oversight" when prompted.

3. Successful AI Recommendations

Despite the logistical failures, the AI proved effective in curating the "experience" portion of the trip:

• Dining: The AI successfully identified a restaurant with high-quality cliff and sea views.
• Activity Planning: The AI effectively mapped out a hiking route along the cliffs and identified a relevant museum in a neighboring town, demonstrating competence in itinerary curation and destination discovery.

4. Methodological Framework for AI Travel

Based on the experience, the following framework is suggested for using AI in travel:

1. Use AI for Discovery: Leverage the AI to find "hidden gems" or destinations that offer a high price-to-beauty ratio.
2. Verify Real-Time Data: Never rely on AI for train schedules, walking directions, or current business operating hours. Always cross-reference with official sources (e.g., Google Maps, official transit websites).
3. Maintain Flexibility: Treat the AI itinerary as a suggestion rather than a rigid mandate. Leave room for "on-the-spot finds" that the AI may have missed.
4. Human Oversight: The traveler must act as the final decision-maker, especially regarding safety and comfort.

5. Synthesis and Conclusion

The experiment concludes that while AI is a powerful tool for inspiration and itinerary building, it is currently unreliable as a standalone travel agent. The primary takeaway is that AI lacks the "situational awareness" required for seamless travel. The traveler’s final verdict—"I would do it again, but I wouldn't rely on AI for real-time information"—underscores the necessity of a hybrid approach where AI provides the creative framework, but human judgment manages the execution.