AI's Initial Promise
The emergence of generative AI has spurred widespread curiosity about its potential to revolutionize various tasks, including travel planning. With two
decades of experience in crafting vacation itineraries, the author decided to put these advanced language models to the test. To gauge their capabilities without risking an ambitious global journey, the focus was placed on a familiar destination: the Finger Lakes region of New York, a place with which the author has a long-standing personal connection. This 'backyard' planning scenario seemed like an ideal way to assess AI's utility for more common travel needs, as opposed to highly specialized trips like African safaris, which might still warrant consulting human experts. The goal was to see if these AI tools could indeed outperform traditional planning methods for a weekend getaway.
First AI Attempts
To begin the evaluation, the author posed a specific prompt to both Google Gemini and OpenAI's ChatGPT, requesting a three-night itinerary for the Finger Lakes in July, focusing on hiking, boating, wine tasting, and exploring small towns, with a rental car for transportation. Both AI platforms generated serviceable plans, consistently recommending Watkins Glen as a central accommodation point, a sound suggestion given its tourism prominence. However, upon closer inspection, significant deficiencies became apparent. Gemini inaccurately suggested Dano's Heuriger for dinner, a restaurant that had been closed since 2024. ChatGPT, while correctly identifying Seneca Lake as home to numerous wineries, only proposed three options, none of which were the author's preferred choices. The information provided, even when accurate, often felt superficial, lacking the depth needed for robust trip planning.
AI's Limitations Surface
Despite the initial shortcomings, there were moments where the AI tools impressed. Both ChatGPT and Gemini rightly highlighted Robert H. Treman State Park, a renowned natural attraction in the Finger Lakes celebrated for its stunning waterfalls and extensive hiking trails. They also suggested staying at the Watkins Glen Harbor Hotel, a well-regarded lakeside property in the area. The author also experimented with Mindtrip, a travel startup leveraging AI for itinerary creation. This platform also delivered a plan that met the basic criteria, including regional highlights and a recommendation for the Argos Inn, a hotel the author frequently suggests. Yet, all three AI services failed to incorporate numerous interesting local spots that the author was aware of and that had been previously featured in Travel + Leisure.
Integrating Expert Insights
To push the AI tools further, a follow-up prompt was introduced, requesting that the itineraries be revised to incorporate recommendations from Travel + Leisure. This directive dramatically improved the quality and interest level of the suggestions. Google Gemini, in particular, explained its adjustments, noting its research into T+L's top-tier picks and 'secret' spots like the village of Aurora and Sheldrake Point Winery's lakeside dining. It also added highly-regarded establishments like F.L.X. Table and 1833 Kitchen & Bar, along with wineries such as Hermann J. Wiemer Vineyard & Winery and Forge Cellars, recognized by the wine press. Mindtrip also responded well, surfacing the Tulip Tree Hostel in Ithaca and highlighting the family-run Glen's Fairway, a popular frozen custard spot in Penn Yan.
Persistent Flaws and Hallucinations
Even with the enhanced suggestions, the AI-generated plans continued to exhibit flaws. A significant issue was the oversight of practical travel logistics; the itineraries often involved excessive driving, a factor the author personally sought to minimize. The AI platforms were also initially sparse on granular details, such as specific wine varietals to try or recommended dishes. More alarmingly, ChatGPT produced factual errors, suggesting pairing filet mignon with dry Riesling and recommending a 'Sri Lankan curry' for brunch at an establishment that did not offer it. These 'hallucinations,' alongside the logistical oversights, reinforced the understanding that while AI can offer a starting point, it still requires critical human oversight and personal refinement to create a truly optimal travel experience.
