Hikers who plan with chatbots are getting confidently wrong advice, rescuers say

Three novice hikers had to be rescued from Mount Shasta in California after relying on Google's Gemini to plan their climb, according to the Siskiyou County Sheriff's Office. One of the men told a deputy the group had leaned heavily on the chatbot for route information and packing advice. The sheriff's office said Gemini had advised them to bring far less food and water than the group needed. They reached the summit at 7 p.m., started down in the dark, got lost, and one of them fell and injured his knee. A planned eight-hour ascent became a multi-day ordeal that ended with a sheriff's search and rescue team and US Forest Service climbing rangers bringing them down.

Wired, which reports the three were 20 years old, uses the case to open a wider look at how chatbots are entering backcountry planning. Rescuers describe a pattern: people who did some research, but handed the judgment to a tool that sounded more certain than it had reason to be.

What rescuers are seeing

"We get plenty of people who come unprepared or don't do any research at all. To these guys' credit, they did try," Nick Meyers, US Forest Service program manager of the Mount Shasta Wilderness Climbing and Avalanche Center, told Wired. "But I think they kind of turned their brains off."

Meyers described other encounters. One Mount Shasta visitor pitched her tent near a "No Camping" sign and insisted she was allowed to because AI had told her so. His team also met a man at a trailhead holding an AI-generated map that placed prominent features on the wrong side of a nearby mountain.

In British Columbia, a helicopter crew rescued a 16-year-old hiker from a ledge in the Widowmaker Arete on the back of Crown Mountain. The teenager had used Anthropic's Claude to help plan the trip. Coverage focused on that detail spread widely in Canada and the US, though Wired notes the hiker later said he had also consulted Google Maps and the hiking app AllTrails.

The Siskiyou County sheriff, Jeremiah Larue, summarised the problem in local reporting: "It tends to want to give you favorable information." He added: "AI can be a good tool, but it shouldn't be used as the expert. Go talk to actual people."

What the numbers show so far

Use is rising. Grand Teton National Park superintendent Chip Jenkins told Wired that the share of visitors who report using AI for travel information "went from zero to 26 percent" over the past year. "And I'm guessing that curve is not going to flatten."

There is no sign yet of a matching rise in rescues. Meyers said his office, which covers a large wilderness area in Northern California, averages about a dozen rescues a year, and 2026 looks like an average year. In Wyoming, search and rescue calls are down: 312 so far in 2026, against 466 for all of 2025 and 377 in 2024. Meyers said it remains to be seen whether more AI use will lead to more accidents.

Some search and rescue veterans see chatbots as the latest in a series. About a decade ago, hikers began taking destinations and information from Instagram and other social media, sometimes with deadly results. Before that, people got into trouble relying on Google Maps or on having phone signal. Jim Moss, a Colorado recreation attorney and search and rescue volunteer, said media have pushed for two decades to get more people outdoors, and it is still unclear how questionable advice from a chatbot will affect them.

Why chatbot advice is a different kind of risk

Chris Flathmann, a human-centered computing professor at Clemson University who has studied AI and search and rescue, told Wired that chatbots carry two specific risks. They can hallucinate, presenting false information persuasively. And their use can lead to what he calls "cognitive surrender."

"When humans work with AI systems, they can get so deep into it they give up thinking critically for themselves," he said. When a chatbot presents itself as an expert on a topic the user does not know, "you can't call BS on that because you don't have the expertise." Once critical thinking stops, he said, it is hard to restart, especially under stress in the backcountry.

Flathmann and others also note that chatbots tend to tell users what they want to hear, with the same authority whether their sources are reliable or not. A detailed set of directions may be drawn from Reddit threads or outdated YouTube videos, or be fabricated outright. The difference from social media, Flathmann said, is the format. TikTok and Instagram offer inspiration. A chatbot offers instructions, right or wrong, along with encouragement. "AI isn't trying to trick people into going up mountains, but I think it's the natural evolution of what happens when you get a tool like this."

The advice from experts has not changed. Check plans against current maps and websites from public-land agencies, trail associations and guidebooks. Better still, talk to a guide, a forest official or a ranger. The Siskiyou County sheriff's office adds practical rules: turn back by noon even without reaching the summit, and carry more than one way to navigate. "If you're going to go into places that will kill you," Meyers said, "you should take your planning seriously."

What it means for work

Trip planning used to mean reading agency pages, guidebooks and trip reports, and weighing them against each other. A chatbot collapses that work into one fluent answer, and the weighing step is the part that tends to disappear. The tasks it removes are cheap to automate but costly to get wrong: estimating food and water, timing a descent, reading a map against terrain. Rangers and search and rescue teams absorb the consequences, and part of their job is shifting toward correcting AI-generated plans at trailheads. What remains open is whether rising AI use will show up in rescue statistics, which so far it has not, and whether chatbots can learn to flag their own uncertainty when the stakes are physical.