Artificial intelligence is becoming part of everyday trip planning. Hikers can ask an AI tool to suggest a route, estimate how long a walk might take, explain the terrain, recommend equipment or describe conditions they might encounter. Within seconds, it can assemble information that might previously have required searching through maps, park websites, guidebooks, weather services and hiking reports.
That convenience can be useful. I use AI regularly myself, including for research, but over time I have become much more conscious of the difference between using AI to help investigate something and allowing it to become the authority on which a decision is based.
A recent High Country News article, AI gets people into trouble in the backcountry, brought that distinction into sharp focus. It described three young hikers who needed help on Mount Shasta in Northern California after a series of planning and decision-making problems. They had not brought enough food and had overestimated their fitness. Instead of turning back when they had not reached the top by late afternoon, they kept going and did not summit until about 7 pm. They descended in the dark, and by the next morning they were lost and one of them had injured his knee. The hikers later acknowledged that some of their preparation and planning had relied on Google’s Gemini AI chatbot.
The article described other examples. One visitor camped near a “No Camping” sign because an AI tool had told her she could camp there. Another person arrived at a trailhead with an AI-generated map that placed prominent landscape features on the wrong side of a mountain. In British Columbia, a 16-year-old rescued from Crown Mountain had used Claude to help plan his trip, although he later said he had also used Google Maps and AllTrails. The article is careful not to claim that AI is driving an increase in rescues. The ranger quoted says it remains to be seen, and Mount Shasta is having an average year. These incidents do raise an important question about what happens when people place too much trust in information delivered by a system that appears knowledgeable and confident.
Cognitive surrender
One phrase in the article caught my attention: cognitive surrender.
Chris Flathmann, a human-centred computing professor at Clemson University who has studied AI and search and rescue, used the term to describe how people working with AI can become so reliant on it that they stop thinking critically for themselves. The problem is harder to recognise when the subject is unfamiliar, because the user may not know enough to spot an error in the first place.
The idea is also being studied more formally. In a working paper, researchers Steven Shaw and Gideon Nave describe cognitive surrender as accepting AI-generated information with little scrutiny and allowing it to replace our own reasoning. In their experiments, participants could choose to ask an AI assistant for help with reasoning problems, and they did so on more than half the questions. The researchers secretly controlled whether the AI gave right or wrong answers. Compared with people working without AI, accuracy rose by 25 percentage points when the AI was right and fell by 15 points when it was wrong. Using the AI also made people more confident in their answers, even when those answers were wrong.
I recognise that behaviour from my own early experience with AI. When I first began using these tools, I placed far more trust in their answers than I do today. Much of the public discussion around AI presented it as extraordinarily intelligent and capable of doing work previously done by people. Coming to the technology without a detailed understanding of how large language models worked, I assumed that information presented confidently and in considerable detail had probably been researched and verified.
As I began using AI more extensively, including testing it against subjects I knew well through my day job and other work, I started finding errors. Some were obvious because I already knew the subject. At other times, something in an answer did not seem quite right, and further investigation showed that information had been misunderstood, assumed, taken out of context or fabricated. References could be misrepresented, and plausible explanations could be built around something that had never actually been established.
That changed how I use AI. I still find it extremely useful for surfacing information, exploring questions and identifying avenues of research, but I no longer assume an answer is reliable simply because it sounds convincing. When factual accuracy matters, I compare the output with what I already know, follow the evidence back to primary or authoritative sources and question anything that does not fit. I have also learned that the quality of the result depends heavily on how I use the tool. If I do not give it enough context, ask the right questions or challenge the assumptions in its response, the answer can look complete while missing the evidence I actually need.
Why this matters in hiking
Hiking is an environment where misplaced confidence can have serious consequences. Planning requires reliable information about the route, weather, current conditions, access and whether the hike suits the people attempting it. Some of that information changes quickly, while other parts depend heavily on local conditions and individual capability.
Someone unfamiliar with an area may have little chance of recognising that an AI-generated answer is wrong. If an AI tool incorrectly describes a route I have walked many times, I may notice immediately. Someone planning their first visit may not. A detailed description of a nonexistent track junction, an incorrect campsite location or an unrealistic walking time can sound credible when the person reading it has nothing to check it against.
This problem is not limited to AI. Guidebooks become outdated, websites contain errors and social media posts leave out important details. Community-generated hiking platforms can hold old or inaccurate route information, and a GPS file is not reliable simply because someone has uploaded it. Another hiker’s recommendation may also be wrong for someone with different fitness, experience or equipment.
AI adds another layer to that mix, and it has a characteristic that deserves attention. It can combine information from many places and present the result as a clear, confident explanation without showing where individual claims came from, how current they are or whether conflicting information exists. The answer can appear more certain than the evidence behind it.
That matters when the confidence of the tool begins to shape the confidence of the person using it. If AI says water is available at a particular location, someone may decide how much to carry based on that claim. If it says camping is permitted, the user may see no reason to check the land manager’s regulations. The risk increases when information is accepted without asking where it came from or whether a more authoritative source supports it.
When the plan stops matching reality
The Mount Shasta incident also shows why this issue does not end once the planning is complete.
The group had still not reached the summit by late afternoon. Whatever information shaped their expectations before the hike, their actual progress was now telling them something important. They were taking longer than expected, daylight was running out and the plan they had started with no longer matched what was happening on the mountain.
This connects directly with the ExploreSafe principle of making decisions early. A hiking plan needs to be reassessed as the trip unfolds. Progress may be slower than expected, a river may be higher than anticipated, weather can deteriorate, terrain can prove harder than described or someone in the group may be struggling. Those observations are new information, and they should influence what happens next.
If an AI tool, website, app or guidebook says a route should take six hours, but four hours in you are nowhere near where you expected to be, the original estimate should no longer control the decision. What is happening around you is now more relevant to the decision than the estimate you started with.
Cognitive surrender becomes relevant when people keep relying on the original advice despite evidence that it no longer fits. Good hiking judgement includes being prepared to question the assumptions behind a plan and change course when those assumptions prove wrong.
Keeping judgement engaged
None of this means hikers should avoid AI. These tools can be valuable during trip research. They can explain unfamiliar terminology, identify questions worth investigating, organise information, suggest things a hiker may not have considered and help locate sources that might otherwise be missed.
Important information still needs to be checked. Track closures and access requirements should be confirmed with the responsible land manager. Weather should be checked through the Bureau of Meteorology or another relevant official service. Fire danger ratings, warnings and restrictions should come from the agencies that issue them. Route information can be compared across current maps, land manager information and reliable hiking resources, while reports from other hikers can add useful context about recent conditions.
The same principle applies wherever the original information came from. If a decision matters, it is worth knowing who produced the information, how current it is and whether another reliable source supports it. That is particularly important when a source is telling us something we want to hear, such as confirming that a route is achievable, conditions should be fine or the trip can go ahead as planned. Flathmann also noted that chatbots can reinforce what users want to hear and present those answers with considerable authority.
For practical guidance on doing this before heading out, see How to Verify Route Information Before a Hike. It explains how to cross-check route details, access information and current conditions rather than relying on a single source.
AI will almost certainly become a larger part of how people research and plan outdoor activities. Used carefully, it can make useful information easier to find and help hikers consider things they might otherwise overlook. The important part is staying involved in the reasoning and continuing to question the information as circumstances change. Hiking decisions still need to account for current conditions, reliable evidence, individual capability and what is actually happening on the ground.
This essay is part of the Human Factors in Hiking series, exploring behaviour, awareness and decision-making on the trail. Explore the series →
