一位记者通过对比测试，评估了AI在旅游规划中相对于人类专家的表现[1]。测试涉及ChatGPT和Claude两个AI助手，针对意大利多洛米蒂地区等地的旅游建议进行对比[1]。
AI在多个方面表现出明显优势。在构建自助导览、比较目的地、推荐住宿和优化路线等任务上，AI展现了突出能力[1]。然而，这些工具也暴露出重大缺陷[1]。AI可能忽视预订需求、低估交通耗时，甚至虚构不存在的景点[1]。塔斯马尼亚当地酒店老板Kristy Probert曾遭遇此类问题——AI为该地虚构了一处名为Weldborough Hot Springs的温泉，吸引了游客前往寻找[1]。
在模仿特定风格或声音方面，AI的表现存在显著差异[1]。Claude能够以Hunter S. Thompson的笔调撰写旅游评论，但在复制旅游记者Michael Gebicki的声音时效果平庸[1]。作者的结论是，AI最适合作为旅游规划的初步工具，帮助制定草案方案，但任何AI生成的建议都需经过人工验证和补充才能付诸实行[1]。
Artificial intelligence systems excel at constructing self-guided itineraries, comparing travel destinations, recommending accommodations, and optimizing routes, according to a comparative analysis conducted by a travel writer [1]. However, the testing also uncovered critical limitations that undermine AI's reliability as a standalone travel planning tool.
When pitted against human expertise, AI assistants like ChatGPT and Claude produced similar suggestions for Italy's Dolomite region but demonstrated troubling weaknesses [1]. The systems have been known to fabricate attractions that do not exist—a problem exemplified by the case of Kristy Probert, a local hotel owner in Tasmania, who reported that tourists arrived seeking the Weldborough Hot Springs, a completely fictional location that AI had invented [1]. Beyond this alarming tendency to hallucinate landmarks, AI tools frequently overlook essential booking requirements, underestimate travel times between locations, and struggle to capture the nuanced, on-the-ground experience that distinguishes authentic travel advice [1].
While Claude demonstrated impressive versatility by writing travel commentary in the style of Hunter S. Thompson, its ability to emulate the voice of travel journalist Michael Gebicki fell flat [1]. These findings suggest that AI functions best as a preliminary planning assistant to draft initial itineraries, but human verification and supplementation remain indispensable for producing trustworthy travel guidance [1].