Travelers Rely on AI for Trip Planning Amid Hallucinations and Trust Issues

A key challenge faced by AI tools is accuracy. These systems, which rely on large language models, often produce ‘hallucinations’—instances where incorrect information is given as though it were factual.

Shyn Yee Ho, a director at tourism consultancy Horwath HTL, reports that her personal encounters with AI have been largely free from such inaccuracies. As a “heavy user” of large language models, she successfully used ChatGPT for hotel recommendations that aligned with her budget and personal interests while planning a six-month solo sabbatical trip. According to Ho, the suggestions were “very clear and good.”

Yet, despite her positive experiences, many travelers remain wary of AI’s reliability.

Rowan recounted an incident where a client, traveling on business in Paris, was late for an appointment after ChatGPT suggested a route that ignored road closures caused by construction. A journey that should have taken ten minutes turned into a 45-minute ordeal.

“Such errors may seem rare, but they actually occur quite frequently,” Rowan noted.

Experts believe that AI has the potential to reshape travel destinations. Smaller establishments may struggle to compete due to insufficient expertise or online visibility.

Independent and older properties, particularly those in developing nations, “will find it even more challenging because they lack the necessary guidance,” Ho stated. “It’s unfortunate, as these locations might need tourism more than ever,” she added.

Conversely, well-known tourist spots could face the negative effects of overtourism as AI’s recommendations continue to favor popular destinations. Guy Llewellyn, an assistant professor at EHL Hospitality Business School Singapore, emphasized that AI should be utilized to highlight lesser-known locations, effectively distributing tourist demand and mitigating the risk of overtourism. However, many AI systems are currently being trained mostly on “top ten” lists, which creates a paradox, he noted.

AI tools may also struggle with real-world complexities that experienced travel agents manage adeptly. For instance, they may fail to consider seasonal weather for outdoor activities or the fatigue from long-haul flights, Rowan explained.

Moreover, AI lacks the capability to assist travelers when unforeseen events occur. For example, during the recent crisis in the Middle East, many travelers found themselves stranded when airspace was shut down. In such situations, AI cannot prioritize individuals for repatriation flights, Rowan noted.

Nonetheless, there is optimism among experts as AI models continue to advance and businesses gain a better understanding of how to integrate these tools. Llewellyn anticipates that as AI becomes more integrated, it may offer more up-to-date information than traditional travel agents.

He stresses the importance of the hospitality industry in structuring and making data accessible for AI systems. “This data doesn’t have to be customer-facing; it could operate purely on the backend through an API call,” he explained, referring to application programming interfaces that allow external access to a company’s data for various applications.

With access to accurate data, AI could minimize hallucinations and enhance the user experience.

Although existing models have their shortcomings, the hospitality sector is actively pursuing the implementation of even more advanced tools. For instance, Booking.com is rolling out AI solutions in collaboration with OpenAI.

Llewellyn predicts that as digitization continues to expand, more AI companies will enter the market, ultimately benefiting travelers by reducing inaccuracies and enhancing overall experiences.

“AI planners are on the horizon,” Llewellyn remarked. “The initial phases may have challenges, but they will eventually bring significant benefits to the industry.”