In a special holiday edition of the BI Tech Memo's "AI Playground," artificial intelligence chatbots, specifically ChatGPT and Anthropic's Claude 4, were put to the test as personal travel assistants for a Tahitian vacation. The experiment involved seeking recommendations for a seven-day trip to Moorea, planned for four couples aged between 50 and 60, celebrating a friend named Theresa's birthday. The user requested comprehensive suggestions, including daytime and evening activities, restaurant and bar recommendations, and the ideal event and location for Theresa's birthday celebration.

Upon receiving the AI-generated itineraries, Theresa, the birthday celebrant, and another friend, Lisa, who had already invested significant time in planning the vacation, provided their expert reviews. Their insights highlighted both the capabilities and limitations of the AI tools.

Theresa noted that both ChatGPT and Claude offered similar activity recommendations, such as cultural tours, 4x4 rentals, lagoon cruises with snorkeling, and a unique suggestion of a sunset cruise for her birthday, which she hadn't considered. In terms of dining, ChatGPT's recommendations proved highly successful, with Rudy's, Moorea Beach Cafe, and the Manava Polynesian show all being booked by the group. Claude also contributed one successful booking with Cocobeach. However, both chatbots suggested Holy Steak House, which was deemed impractical due to its 40-minute taxi ride from their hotel. Theresa expressed a strong preference for ChatGPT's day-by-day itinerary format, finding Claude's approach to be overly focused on marketing for the Cook's Bay hotel.

Lisa's assessment further elaborated on the differences. She found ChatGPT's response to be more comprehensive, featuring a sample daily itinerary complete with pricing estimates and source/reference links. ChatGPT provided a wider array of options and presented its suggestions in an easy-to-read bullet format, though Lisa found the accompanying icons somewhat "gimmicky or overused." While Claude's response was visually appealing and easier to read, it offered a noticeably smaller selection of activities and dining choices. A significant oversight noted by Lisa was that neither AI bot mentioned scuba diving as a possibility, despite Moorea being an excellent diving destination and many in their group planning to participate in the activity. Overall, Lisa rated ChatGPT 4.5 out of 5 stars, while Claude received 3.5 stars.

The experiment underscored the potential of AI chatbots in assisting with travel planning, providing valuable suggestions and organized itineraries. However, it also highlighted the importance of human input and specialized knowledge to refine recommendations, ensure practicality, and include niche activities that might be overlooked by general-purpose AI models.