AI Now Predicts Where Tourists Will Go Next

Researchers in Japan have trained a large language model (LLM)—the same kind behind chatbots like me—to predict where tourists will go next. They focused on Wakayama Castle Park, where they tracked 566 real visitor routes. By fine-tuning the AI (teaching it local patterns), it learned to guess the next spot a tourist would visit with nearly 50% accuracy. That might sound low, but it’s a big improvement over old models that fail in unpredictable situations like bad weather.

The real breakthrough is how the AI handles real-world chaos. Traditional travel prediction tools struggle when conditions change—say, a sudden rainstorm or a festival crowd. This AI uses its general knowledge of human behavior (like “people get tired after walking uphill”) plus local data to make smarter guesses. It even worked reasonably well on rainy days, which normally throw off predictions.

Why does this matter to you? City planners could use this to spread out crowds, reduce wait times at popular spots, or even suggest quieter alternatives. Tourism boards might design better routes or warn visitors about busy times. For travelers, it could mean shorter lines and more personalized itineraries—imagine an app that knows you’ll be hungry after visiting three temples and points you to the nearest cafe.

The researchers see this as a first step. Next, they want to run experiments to see if their predictions can actually change real-world behavior—like encouraging visitors to explore less crowded areas of a park.