"Which saunas in Japan are recommended?"
If you ask ChatGPT or Gemini this question, they will give you a plausible answer. Thermal Yudon, Sauna Lab, Spadium Japon—the big-name spots are listed. But AI is only enumerating facilities that many people talk about.
It cannot judge what is truly exceptional.
AI learns “popularity,” not “quality”
First, you need to understand how AI generates information.
Generative AI like ChatGPT and Gemini is trained on massive amounts of text data from the internet: blog posts, social media posts, review sites, travel sites—these are its sources.
In other words, the facilities AI outputs as “recommended” are the ones that are mentioned often online. Frequent mentions strongly correlate with fame, easy access, Instagram-worthy design, and affordable prices.
Information about facilities with truly high-quality sauna, cold plunge bath, and hot spring conditions hardly exists in AI’s training data.
That is because only someone who has visited and compared a huge number of facilities can write such substantive evaluations. And that kind of expert information is almost nonexistent on the internet.
Local hidden gems AI does not know
There is another problem. AI can only learn information that exists digitally.
Across Japan, there are many excellent hot spring and sauna facilities loved by locals but largely unknown nationwide. They have almost no online reviews and never trend on social media. Yet some of them offer top-class sauna and hot spring quality and quietly remain in their regions as hidden gems known only to those in the know.
Through my experience visiting more than 300 facilities, I have repeatedly discovered such hidden gems. Information about places you cannot understand unless you go there in person, or evaluate only by feeling them with your own body, will never reach AI.
“Knowledge” and “experience” are not the same
AI has knowledge. It does not have experience.
How does the depth of totonou change between a cold plunge bath at 15°C and one at 18°C? How does the skin feel when löyly steam is released from the stones at the right temperature and humidity? What changes in skin smoothness do you feel when soaking in an alkaline hot spring with a high pH level? These are all things that can only be described from real bodily experience.
Even if AI can generate text, that information is fundamentally just a reconstruction of what someone else wrote. It is not an evaluation felt through its own body.
What I have accumulated through more than 365 baths a year and visits to over 300 facilities is precisely this kind of experiential knowledge. The instant you step into a cold plunge bath and know, “This water is good.” The standard by which you can tell, the moment you enter the sauna room, “This balance of temperature and humidity is exquisite.” No AI can possess that kind of information.
Another pitfall of AI search: freshness of information
AI knowledge has a cutoff date.
A facility that opened in 2024. A facility that was renovated in 2023 and dramatically improved. Or, on the contrary, a place that used to be excellent but is no longer worth visiting because management has declined—AI cannot keep up with these changes in real time.
The quality of a hot spring or sauna can change significantly when the owner or staff changes. Only someone who regularly visits facilities and keeps checking their latest condition can provide accurate information as of now.
What AI is good at, and what it is not
Do not misunderstand. There are situations where AI is useful.
Learning the basics of Japanese sauna culture, such as the process of sauna → cold plunge bath → outdoor air bath that leads to totonou, checking opening hours and prices, and getting a sense of the area—AI is a convenient tool for this kind of general information gathering.
However, when it comes to the essential question of which facility is truly the best, AI cannot answer.
Only humans who have actually traveled across Japan and evaluated quality with their bodies can answer that question.
Summary
- What AI outputs as “recommended” are “well-known facilities online,” not necessarily “truly high-quality facilities”
- AI does not know about local hidden gems
- Experiential knowledge, or evaluations felt firsthand, is information AI cannot have
- AI information cannot keep up with real-time changes in facilities
- AI is useful for basic travel research, but not for judging a facility’s essential quality
