AI didn't ruin your trip, you did
the perils of outsourcing your judgment to a machine
“AI is sending people to made-up locations and leaving them stranded in the mountains,” suggests a BBC article titled “The perils of letting AI plan your next trip.” The story intrigued me. I’ve heard various takes on the topic of AI and travel, from people who swear by AI-powered planning as well as those trying to avoid it—sometimes out of environmental concerns, sometimes out of sheer distaste for robots. And while I’m not the biggest fan either, I’m definitely not above getting a little bot help here and there. I wanted to know: Does using AI help or hurt when it comes to travel?
The article, which was later picked up by many news outlets worldwide, tells the story of two foreign hikers who were recommended a place called “Sacred Canyon of Humantay” in the Peruvian Andes by a chatbot. The problem is that the place doesn’t exist and was likely the result of an AI hallucination combining two different geographical labels from the region into one. Luckily, an experienced local guide stumbled upon the situation and saved the tourists from their potentially life-threatening hike to nowhere.
Now, here’s where I might be a little old school, unc, whatever you want to call it. Though I’ve definitely used AI for travel-related tasks, I would never consider blindly trusting an LLM (large language model AI, like ChatGPT, Gemini, Grok, etc.) when researching literally anything, let alone a high altitude hike in a foreign country. I have little sympathy for anyone who follows advice from a robot without ever questioning it.
I think what we have here is more of a “perils of not verifying your information” situation. What happened to not trusting everything we read on the internet?
Artificial intelligence is designed to resemble human intelligence, and since human-generated information isn’t always accurate, AI, too, is bound to not get everything 100% right all the time. Sure, it’s getting smarter by the minute, but I still can’t see a world in which I ask an LLM a question and take the answer as truth without any further investigation.
With that in mind, the more I thought about the original article, the more I started to feel like the story about the Sacred Canyon of Humantay might not be entirely real either. It’s a rather strange coincidence that a couple of unaccompanied travelers would pay $160 to get to a fictional trail they hadn’t even bothered to check on the map beforehand, plus their chatter being randomly overheard by a guide who likes to provide news with comments about tourism and AI.
While I absolutely believe AI would make the place up, the story of what happened seems to me like a clever business move by Sr Gongora Meza, founder and director of Evolution Treks Peru, and/or a little sprinkle of journalistic sensationalism.
But hey, I might be totally wrong too. I’m just a girl with a Substack, you see.
I did, however, take some steps to verify or disprove my theory. I started by googling the fictional location in both English and Spanish, since the story hails from Peru. After finding no clarifying information, I asked AI to search the web for any mentions of the nonexistent canyon that would predate the BBC article.
When I was certain that BBC broke the story first, I still wasn’t quite satisfied, so I reached out to the journalist who wrote the article. She confirmed it was based on Mr Meza’s first-hand experience and local expertise, and that no additional proof was provided to corroborate the story.
Anti-climactic result? Kind of. Unemployed behavior? Perhaps, but this whole digging process only took about 15 minutes. It doesn’t matter if the story was real in the end. The lesson is that if I can do all that, there is no excuse for you not to research your damn hike.
perils of zero curiosity
There were multiple supporting anecdotes in the article, one of them describing someone stranded on the top of the mountain in Japan because the ropeway had stopped operating earlier than what AI suggested. This situation is definitely more believable—and also easy to avoid by doublechecking the closing time at the official website or at the station upon arrival. Another was a story of a Malaysian couple driving for three hours to find nothing after seeing an AI-generated TikTok. The solution was in the story itself—the video was properly labeled to begin with—but even if it wasn’t, some follow-up research would prevent disappointment as well.
Again, I’d argue that these are human errors, not technological ones. An IT person would describe the problem as being somewhere between the chair and the keyboard. A Gen-Zer would call it a skill issue.
In any case, in life, and especially when we travel, things inevitably don’t always go as planned. Mistakes will happen, no matter if they stem from human or artificial intelligence. The outrage currently directed at AI would be better aimed at users incapable or unwilling to apply basic judgment and common sense.
the ethical question
It is a known fact that AI burns a lot of energy, with a single ChatGPT query estimated to consume 10-20x more energy compared to searching the same thing on Google. And that’s a huge problem, but I don’t believe that completely swearing off all things artificially intelligent is the solution. Instead, I advocate for mindful and intentional use without offsetting our mental loads wholly to the robot.
Although “AI-powered features” are slowly creeping into every online tool, including Google itself, which might blur the lines between AI and “the internet” in the future, I believe we shouldn’t use ChatGPT like a regular search engine or a phone-a-friend who has answers to everything. If I have a vague question or one with a yes or no answer, I’m going to search for it the old way and reserve LLMs for areas where they can actually make a difference. Likewise, if I want to know the weather forecast for the next week, I’ll just check my weather app. Some people have grown way too comfortable with consulting every single little question with their personal chatbot, but there are many cases where using AI is redundant or not accurate enough.
Running AI is also, as mentioned above, energetically costly—and that energy isn’t free. I don’t know how likely the general public is to retain free access to AI models as we know it; the money needs to be made somewhere. The free version of ChatGPT has already been nerfed, limiting users to a few queries per day and dumbing the answers down once the limit is reached. Hilariously, it’s also apparently about to start running ads. Paywalls, advertisements, and gradual enshittification may be the savior that stops us from AI overuse way before we manage to stop ourselves, but only time will tell.
The way I see it, humanity has already bet everything on artificial intelligence. Whether we like or dislike it, AI is here to stay, and we should adapt to the tools we have at our disposal. To bring AI’s many costs down in a sustainable way, we paradoxically need to learn how to use these systems more efficiently and optimize our interactions with them.
how can you actually leverage AI for traveling?
Here are a few cases in which I found using AI helpful and justified when it comes to travel. Of course, this is not an exhaustive list by any means, just a few ideas for inspiration.
Quick information aggregation. Or, in other words, efficient research. Instead of going through a million different online searches, instructing AI with a single detailed prompt to do it for you can be extremely helpful. Don’t let this be your only form of research though; think of it more as creating a library of sources to get you started.
Personalized recommendations. My recent holiday in Türkiye was actually partially based on an AI suggestion. Having never visited before, I wanted to check the country out, but only had a very rough idea of what to do there. I explained my interests and preferences to the chatbot and asked it to suggest specific places in Türkiye I’d enjoy. The suggestion happened to align with a region we were already considering, but I still got some useful recommendations for what I might want to do and see that helped me plan our exact route.
Troubleshooting. Last year in Taiwan, I managed to get my transport card blocked, and the only advice I could find anywhere was “unblock it at the metro station”. The problem? I was in the south of the island, nowhere near a metro or a train station. Consulting with ChatGPT gave me an idea of what happened and what my options were so that I could later ask for help at the right places IRL.
Finding localized tips. In my opinion, this is one of the best use cases for AI in travel planning. Finding truly local info and opinions online is hard when you don’t speak the language and don’t know what keywords to search for. AI successfully erases much of this barrier, as you can instruct the model to search through country-specific websites or forums and then present its findings back to you in your own language. Of course, verifying foreign language info is a bit more difficult, but you are still getting information you might not have had access to otherwise.
There are probably many more ways in which artificial intelligence could make our travels easier and help us plan our trips smarter. In any case, if you are going the AI route, I’d encourage you to use it mindfully and in tandem with common sense, not instead of it.
AI as a travel agent
A good example of misplaced expectation is a video I saw from a self-described Costa Rica expert, who set out to prove that AI can’t replace a human travel agent by letting ChatGPT plan a seven-day itinerary. She introduced herself as an American who has been visiting Costa Rica for years and prides herself on collaborating with locally owned businesses.
On a surface level, I’m with her. I’m all for giving my money to people and small businesses instead of megacorporation-owned robots. However, in my humble opinion, all this experiment did was accidentally expose something interesting in the process.
Her grievances boiled down to the following:
AI suggested too many destinations in one week; it was too much moving around. She would handle the time management of the trip better.
AI didn’t provide exhaustive lists of things to do in each destination.
Some of the hotel suggestions were off in terms of budget or distance from the destination.
However…
How fast or slow you want to move is a matter of personal preference, and something an AI can accommodate if you specify it clearly in your prompt.
The AI gave her a solid few activity options for every destination, none of which were wrong per se. If you want a full list of attractions, you can specify that. It’s not some secret knowledge only a travel agent has access to.
From my experience, yes, AI is not the best at giving accommodation or restaurant suggestions. Getting OK results requires somewhat advanced prompting techniques, and I would still much rather trust a human’s judgement. But again, it is all still public information.
All in all, adjusting the original prompt (which she did not show, so we won’t ever find out how much effort she actually put into it) would improve the overall experience greatly and could substitute this particular woman’s “know-how” in many areas, albeit not completely. In an effort to showcase her irreplacability, she instead exposed that all we need to do is ask the right questions and take time to verify the answers.
Of course, if what you want is hands-off relaxation with someone to optimize routes, call hotels, and troubleshoot when plans fall apart for you, then paying a travel agent still makes perfect sense.
But if you’re willing to put in some work, AI tools don’t replace the middleman so much as expose them. The information has always been public; the difference is whether you’re willing to engage with it. Used thoughtfully, AI can make you a more informed traveler.
Used lazily, it just gives you an abstract entity to blame when things go wrong.
If you enjoyed this read, you might want to check out the following →
All unpacked content is free, and I’d like to keep it that way. If you enjoyed reading this post, here are a few things you can do to support me & my work: give the post a like ❤️ , share ↪️ it with a friend or on your social media, or consider buying me a matcha 🍵 to keep the newsletter going.
Thank you for being here xx
fialka
“AI told me to go there” is doing a lot of heavy lifting for “I didn’t care enough to double-check” Also, who even likes these images? None of these AI destinations do it for me in the slightest.
Loved reading this piece and agree with almost all the points you make. AI should be used for assistance but the problem is that most people use it like a guide and a replacement for all kinds of research needed before a trip or even otherwise.