AI vs. Humans: Who Should Plan Your Hike?
Please, I beg you - don’t let ChatGPT plan your hiking trip.
Now, I want to be clear: I’m not against the use of AI in general. I do have strong opinions about how AI will shape our future (and truthfully, it seems a little bleak). That said, I’ll be the first to admit that ChatGPT, and other AI tools like it, are incredibly useful. They are excellent for quick summaries, for creating and rewriting content, for drafting emails, for brainstorming and generating new ideas, and so on. I use it all the time for all of these things and more. Even my mom uses ChatGPT instead of Google search these days.
And I know that lots of people are now using AI as a trip planning tool. But I promise, if you use it to plan your hikes for your trip to Switzerland, you are going to be very disappointed. Why, you ask? Let me show you:
Reason 1: AI is terrible (terrible!) at reading maps.
It’s been reported that ChatGPT isn’t great at math, but I also wanted to put it’s map-reading skills to the test. I asked ChatGPT to simply route a hut-to-hut tour anywhere in the Berner Oberland. While it did get the general region correct, let’s take a look at it’s suggestion for the second day: Hiking from the Lobhorn Hut to the Faulhorn Berghotel.
Ok, this looks great, right? The most scenic trail in the region, it says!
The problem is that all of the hiking statistics are completely wrong. AI’s version makes this hike sound way shorter and easier than it actually is. If you actually route it out using proper hiking apps (my favorite, by the way, is SwissTopo), you’ll find that in reality, this hike is a beast: clocking in at 26 kilometers, 2,000+ meters of elevation gain, and a whopping 10.5 hours of hiking. Yikes!
And by the way: ChatGPT also doesn’t mention is that there are transit options that can potentially make this exact route more manageable - like taking the local gondola from Sulwald to Isenfluh, then a bus to Wilderswil, and finally the mountain train up to Schynige Platte. But here’s the catch: you’d only know about those alternatives if you dug into the details yourself. AI isn’t going to spell that all out for you unless you already know what to ask.
For casual hikers especially, this is where AI can really lead you astray. That gap between “AI-easy” and “real-life-hard” can be the difference between a fun adventure and a pretty miserable experience.
Reason 2: AI will often “hallucinate” and give inaccurate - and potentially dangerous -information
AI models like ChatGPT are infamous for what’s called hallucination. And no, sorry, not the fun kind. It means it will generate information that’s completely wrong, while confidently presenting it as fact. Basically, if it doesn’t know the answer, it will fill in the gaps with something that seems plausible, but isn’t true.
Take this example: I asked ChatGPT for another hut-to-hut suggestion, suitable for all types of hikers. Here’s what it gave me:
Again, looks great on paper, right? A multi-day hike with the famous Oeschinensee and three classic SAC huts - perfect!
But unfortunately this recommendation isn’t just misleading - it’s actually pretty dangerous. It’s pitched as a “family-friendly” hut-to-hut hike, but anyone who has actually done these trails knows that is very far from the reality. Lets take a closer look at the details ChatGPT provides for two days of this hike:
First, the stats for the daily distance and elevation change are way off (the hike up to Bluemlisalp Hut, for example, is shockingly, achingly steep), although honestly, in this case that’s the least of your worries.
The bigger problem is that this route directs hikers onto a trail that has been closed since 2020 due to rockfall danger. The official detour, meanwhile, follows a highly exposed path that clings to the cliffside.
The Fründenhütte itself warns about how tough this hut ascent is, and even recommends using a via ferrata (rock climbing) set for that section. This is far from suitable for families or inexperienced hikers, and taking this advice could put someone in a very serious situation.
And here’s the kicker: ChatGPT even reassures you that there’s an “easier alternative route” if this one feels like too much. Except… there isn’t. That trail doesn’t exist. The Fründenhütte has only one way in and one way out. Yikes again.
Reason 3: It only corrects itself if you call it out. And even then, it might still be wrong.
Another big issue is that AI tools don’t know when it’s wrong. If you catch the mistake and point it out, it will apologize and revise its answer (it’s very humble in that way, at least!). But it relies on you, the user, to realize that it was wrong in the first place.
Here, I asked for a 4-day hike in Valais, and noticed that one of the hut-to-hut was wildly inaccurate.
Chat GPT very kindly apologized, acknowledging that oops, the day hike it suggested actually crossed a couple glaciers and therefore wasn’t a viable route after all. Kind of a big oops, right? But if you’re not intimately familiar with Swiss mountains and trails, it’s nearly impossible to spot that this suggestion doesn’t actually make sense.
Morevoer: even its apology was wrong. It mixed up the mountain chain entirely, and in its revised suggestion, it sent me up Pas de Chèvre - another pass that had been closed for years due to rockfall danger. In other words, the “fix” was just as misleading as the original mistake. I’m sorry ChatGPT, I know you are trying your hardest…but again, yikes!
Reason 4: It’s just better to learn from real people.
Ok, maybe this reason is a little biased 🙂 I promise, I didn’t write this just to completely rag on AI’s mapping skills. But even if ChatGPT and all it’s friends eventually gets better at reading topo maps (and I’m sure it will) and stops it’s fantastical hallucinations, the fact is that maps, timetables, and elevation profiles are only part of the story. Human judgment is another.
A real person can look at the same route and immediately spot things AI won’t: a brutal, knee-busting summit-to-valley descent you might want to skip, a stretch through alpine pastures that tends to turn into a muddy, manure-filled slog after rain, or a narrow mountainside traverse that might be fine for seasoned hikers but terrifying for anyone with a fear of heights. In other words, people don’t just read out the data: they interpret it, filter it, and add experience, context, and common sense.
The takeaway? Combine AI with real expertise.
So go ahead, use AI for the heavy lifting of ideas, summaries, pro/con lists, etc. But for the actually planning - integrate something human. For some, a bit of targeted Googling combined with trusted blogs and resources can go a long way, especially if you already know how to read maps and have some trail experience. But if you don’t know how to read topo maps, are unfamiliar with the region, or just need help turning your idea into a route that actually works, try reaching out to a real person for help. A local, a guide, a hiking friend—or better yet, me (since planning hikes is what I do!). That way, you’ll hit the trail with a plan that’s not just theoretically possible, but smart, practical, and genuinely enjoyable.
Now, let’s just hope that the machines don’t find this blog post in 10 years and come take their revenge.