- High Country News, via Wired, reports rescuers who met hikers that had leaned on AI chatbots for trip planning. One expert says the risk lies in answers users cannot check.
- Rescue numbers do not yet show a rise: the Mount Shasta office calls 2026 an average year. The behavior may still matter for teams using chatbots on unfamiliar topics.
- Ask which decisions in your organisation rely on AI answers, and who checks them against an authoritative source.
Trusting what you cannot check
Every tool gives wrong answers sometimes. The question is what happens when the person using the tool has no way to tell. That is the idea worth taking from a story about hikers in trouble.
The argument here is ours, not the rescuers'. A beginner cannot easily judge an answer that sounds expert. That may be exactly when the beginner leans on it most. The sources do not show this happening in each case. They do show rescuers and a researcher worried about people who stop checking.
What happened on the mountain
High Country News reported the case, and Wired republished it. Three hikers in their twenties called for help from Mount Shasta in Northern California. They had packed too little food. They had overestimated their fitness. They kept climbing in late afternoon and reached the summit at 7:30 pm, then had to come down in the dark.
The hikers admitted later that their poor planning came partly from relying on Google's Gemini chatbot. The report does not say what Gemini told them. By morning they were lost on a rough part of the mountain, and one had a banged-up knee.
A local climbing and avalanche center, run by the US Forest Service, answered the hikers' call. Its program manager, Nick Meyers, said the group at least tried to research the trip. In his words, "they kind of turned their brains off."
Meyers described other cases. One camper pitched a tent near a "No Camping" sign and insisted AI had said she could. At a trailhead, his team met a man carrying an AI-made map that put well-known landmarks on the wrong side of a nearby peak.
The report adds a caution. A 16-year-old rescued in British Columbia had used Anthropic's Claude to plan his trip, and that detail drove viral coverage. The teen later said he also consulted Google Maps and AllTrails. One tool rarely explains an incident by itself.
How an expert says the failure works
Chris Flathmann, a Clemson University researcher of human-computer interaction, has looked at how AI is used in rescue work. He points to two risks. One is hallucination, where a chatbot presents false information as if it were fact. The other is what he calls "cognitive surrender": people working with AI give up thinking critically.
Flathmann and others also say chatbots lean toward telling people what they want to hear, in a confident voice, even when their sources are weak. Such an answer can look like a careful route plan. It may be pieced together from Reddit threads and outdated YouTube videos, or made up.
The plain lesson is that a chatbot produces text that sounds plausible. It does not verify facts as it writes. A confident tone therefore says nothing about whether the answer is right.
He contrasts this with social media. TikTok and Instagram offer inspiration. AI tends to give instructions and encouragement. He adds that AI is not trying to trick anyone. In his view, this is the natural result of such a tool.
Unfamiliar topics make it harder. As Flathmann put it, when AI poses as an expert, you cannot challenge it without the expertise yourself.
What the numbers do and do not show
The evidence on harm is thin so far. Grand Teton National Park's superintendent, Chip Jenkins, said the share of visitors reporting AI use for travel information went "from zero to 26 percent" over the past year.
Rescue counts do not show a matching rise. Meyers' office averages a dozen rescues a year, and 2026 looks average. Wyoming has logged 312 search and rescue calls so far in 2026, against 466 in all of 2025 and 377 in 2024. The 2026 figure covers only part of the year.
Veterans also note a history. Hikers were once led astray by Instagram, by Google Maps and by phone coverage that was not there. Jim Moss, a Colorado recreation attorney and rescue volunteer, said it is still unclear how chatbot advice will affect the larger crowds now outdoors.
What this means for your organisation
The sources say nothing about workplaces. The link is our analogy. Some teams may use chatbots on topics the user cannot judge: a contract clause, a security setting, a tax rule. The structure looks similar. Whether the same habits form there is a question to test, not a finding.
The experts' advice for hikers carries over. Check against authoritative sources, such as agency maps, trail association sites and guidebooks. Better still, talk to a person who knows the ground.
Questions to put to your team:
Which decisions now start with a chatbot answer? Who checks that answer against the primary source, and how? Does anyone ask the chatbot for its sources, then open them? Which topics need a named human expert before anyone acts?
A chatbot can make a plan look finished. The work of checking it still belongs to a person.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Wired.
