On a recent trip to Taiwan, I turned to ChatGPT to ask for recommendations for the best beef noodles in my area — with the very specific request that the shop had to accept credit cards, because I was running low on local currency.
The chatbot recommended a place that was a short walk and featured the most delicious beef tendon I’ve ever had. I was pleased to be the only foreigner in the no-frills, no air-conditioning joint that was home to a fat, orange cat taking a nap under one of the metal stools.
But after my meal, I panicked when the impatient woman behind the counter had to put aside the dumplings she was folding to try and communicate in English to me that it was cash only. Even a quick Google search would’ve saved me from this fate, and I felt foolish for blindly trusting AI’s outputs.
Talking to other travelers, I realized I was lucky the restaurant existed at all, hearing stories of AI tools sending confused tourists to places that were closed or not even real. Still, I found the tool helpful while navigating a foreign city, using it not just to find spots to eat but also to translate menus and signs, as well as communicate with locals via voice mode. It felt like the ultimate Asia travel hack.
People are also reading…
Back home in Tokyo, where a weak yen helped make Japan a top destination for global travelers, I decided to put various AI platforms to the test. I asked DeepSeek, ChatGPT and the agentic tool Manus to create itineraries for someone visiting the city or Japan. The results were jam-packed and impressive, but mostly featured all the same tourist spots that you’d find at the top of sites like Trip Advisor.
Some of the recommendations were also out of date. ChatGPT advised staying in a traditional inn that closed over a year. Even my request for more off-the-beaten-path locations spit out areas I specifically avoid at peak times, like Shimokitazawa, because of the crowds of tourists.
The outputs made sense given these tools are an amalgamation of data scraped from the internet. It does save travelers the step of having to scroll through hundreds of websites themselves and put together an itinerary on their own. But relying on this technology also risks a further homogenization of travel.
Already, the tech industry is being blamed in tourist hotspots for creating feedback loops that push visitors to the same destinations — with winners and losers chosen by a powerful algorithm. Given that AI systems are predominantly trained on English-language text, this also can mean local gems easily slip through the cracks of training data.
I can’t imagine the late Anthony Bourdain eating pho on a stool anywhere in Vietnam that even had a website.
AI isn’t entirely to blame. Before the rise of these tools, social media was already reshaping travel — sometimes in bizarre ways.
An influencer on Chinese social media posted about a railroad crossing in my neighborhood and now it is constantly inundated with people doing photoshoots. One of my favorite summer swimming spots in the outskirts of the city unexpectedly went viral on TikTok last year, and it was shocking to see how crowded the riverbanks had become with foreigners. A town near Mt. Fuji garnered international headlines last year after briefly erecting a barrier to block the view of the iconic landmark when it was overrun with tourists trying to all get the same shot — behind a convenience store, of all places.
Of course, this isn’t limited to Asia. As AI applications proliferate, more people are turning to them to plan vacations from Barcelona to New Orleans. But the data they're trained on has inherent limitations. Perhaps it wouldn’t hurt to put the phone down and ask a local for their top spots.
Ultimately, AI can break down language and cultural barriers for travelers in ways that seemed unimaginable a decade ago. That’s a good thing, and the convenience is undeniable. But it's good to remember that some of the best parts of travel can never be optimized by a machine.
