Three novice climbers recently had to be rescued from Mount Shasta after a summit attempt went badly wrong. They reached the 14,179-foot summit at 7 p.m., descended in the dark, wandered off route into Mud Creek Canyon, and eventually spent the night stranded after one member of the group injured his knee.
There were several bad decisions along the way, including continuing seven hours beyond Mount Shasta’s recommended noon turnaround time. But one detail understandably grabbed everyone’s attention: the climbers told rescuers that they had relied heavily on Google’s Gemini AI to help plan the trip.
According to the Siskiyou County Sheriff’s Office, Gemini helped them choose a route and determine what to bring, and the group said it recommended substantially less food and water than they eventually needed. What they expected to be an eight-hour summit attempt turned into a multi-day ordeal.
AI models can be incredibly useful tools. But I would not let one decide how much water I need to climb a 14,000-foot mountain.
Would you use AI to plan a hike up a 14,000-foot mountain?
Quick Verdict: Should You Use AI to Plan a Hike?
AI can be useful during the early stages of planning a hike. It can help you brainstorm routes, build a preliminary packing list, organize information, or point you toward questions you hadn’t thought to ask.
What it should not be is your final source for current trail conditions, route selection, weather hazards, water availability, permits, navigation, avalanche conditions, wildfire closures, or other information where getting the answer wrong could put you in danger.
Use AI to help conduct research. Don’t confuse it with the research itself.
What Happened on Mount Shasta?
The three climbers established a base camp around 8,400 feet on August 29, then started toward Mount Shasta’s summit at 3 a.m. the following morning. By about 1 p.m., they had only reached 11,000 feet and discussed turning around. They continued instead and eventually reached the summit at approximately 7 p.m.
That was the first major problem.
Mount Shasta climbers are advised to have a firm turnaround time around noon or 1 p.m. on the Clear Creek Route. The Mount Shasta Avalanche Center specifically warns climbers not to let exhaustion or summit ambitions interfere with route finding and emphasizes returning the same way you climbed.
The inexperienced hikers used Google Gemini to plan their trek.
The group descended after dark and called the sheriff’s office for directions about an hour later. They eventually wandered into steep terrain in Mud Creek Canyon, where one of the climbers fell and injured his knee. They spent the night there before Forest Service climbing rangers and search-and-rescue personnel reached them the following morning.
AI wasn’t responsible for every decision those climbers made. That’s important. Gemini didn’t force them to keep climbing after their turnaround time.
But the group told rescuers they had depended heavily on Gemini for information about the route and what they should pack. The sheriff’s office called that reliance a “critical misstep.”
That’s the part that interested me. So I tried something similar myself.
I Asked ChatGPT to Plan a Mount Shasta Climb
I asked ChatGPT to help put together a Mount Shasta summit plan. The result looked fantastic.
As an experiment, I asked ChatGPT to create a hike plan for me for summiting Mount Shasta.
It created a polished infographic showing a multi-day Mount Shasta itinerary, route information, an elevation profile, a gear checklist, permit information, and tips for completing the climb. It was exactly the sort of thing you could screenshot and send to your hiking partners while feeling like you’d accomplished a productive evening of trip planning.
There was just one fairly substantial problem. The infographic routed me up Avalanche Gulch. At the time I ran the experiment, the Mount Shasta Avalanche Center was explicitly saying not to climb Avalanche Gulch because discontinuous snow and significant rockfall had made it a poor late-season route. The center instead identified Clear Creek as the best late-season route for most climbers.
Here’s where things get especially weird. ChatGPT apparently knew that. Immediately beneath the infographic, its own text warned me not to follow the Avalanche Gulch itinerary it had just created. It correctly referred to the August 30 Mount Shasta climbing advisory, explained why Avalanche Gulch was inappropriate under current conditions, and suggested planning around Clear Creek instead.
In this expanded view, you can see that ChatGPT explicitly told me to not follow the hiking plan it had designed for me.
So I had one AI response essentially telling me:
Here’s your route.
Followed immediately by:
Don’t use that route.
That’s a problem when the attractive, easily shareable graphic is the part someone might save to their phone.
It Also Gave Me Outdated Permit Information
There was another small but revealing problem in the generated graphic.
ChatGPT listed the Mount Shasta summit pass as $15 per person, “as of 2024.” The current summit pass is $25.
That’s not going to get somebody killed, obviously. But it demonstrates a larger problem with using generative AI for outdoor trip planning: even when most of an answer looks right, individual pieces of information can be stale, mismatched, or simply wrong.
If AI can’t reliably give me the current permit fee without verification, I’m certainly not going to assume its information about snowpack, water availability, route hazards, or required mountaineering equipment is correct.
The Problem Isn’t That AI Knows Nothing
This is where the conversation gets more interesting than simply saying “AI bad.” The ChatGPT response I received contained plenty of good information. In fact, once I got past the incorrect graphic, it pointed me to exactly the kinds of sources I would recommend: the Mount Shasta Avalanche Center, local trail organizations, and Forest Service resources.
That’s one of the things AI is genuinely good at.
You can ask something like:
“What information should I research before climbing Mount Shasta?”
and get a useful checklist of subjects: weather, route conditions, water sources, equipment, permits, fitness requirements, navigation, turnaround times, emergency planning.
Great. Now you know what you need to investigate.
The mistake is asking:
“Tell me everything I need to know to safely climb Mount Shasta,”
and treating whatever comes back as authoritative.
Those are two very different uses of the technology.
Current Trail Information Has a Short Shelf Life
Hiking information can change quickly. Mountaineering information can change absurdly quickly. That’s especially obvious on Mount Shasta.
When I had ChatGPT make me a hiking plan, the Avalanche Center’s August 30 advisory said Clear Creek was generally the best late-season option, while Avalanche Gulch wasn’t recommended. It also warned that an approaching storm could dramatically change conditions.
A few days later, an update reported that roughly 12 to 18 inches of snow may have fallen above 10,000 feet, substantially changing upper-mountain conditions.
A route recommendation that made sense last week may not make sense this week.
AI systems can search current information, but they can also mix that information with older material, misunderstand dates, summarize sources incorrectly, or—as my Mount Shasta infographic demonstrated—put outdated and current recommendations into the same answer.
A local ranger or avalanche center has another enormous advantage: someone is actually looking at the mountain.
Use trustworthy, local, and regularly updated sources of information for planning your backcountry explorations.
Mount Shasta Isn’t Even a Normal Hiking Trail
There’s another detail from the Mount Shasta Avalanche Center that makes this incident especially instructive. There is no marked or sanctioned trail to the summit.
Lines shown on maps represent suggested climbing routes, and the Avalanche Center explicitly says that route-finding decisions have to be made in the field according to current conditions, experience, and judgment. The upper 3,000 feet can involve rockfall, dangerous slips and falls, extreme wind, and quickly deteriorating visibility.
That’s an important distinction. A line on a screen isn’t necessarily a trail. I think years of using Google Maps and hiking apps have conditioned us to expect that when a digital map shows a line from Point A to Point B, following that line gets us there. Sometimes it does. On a mountain, sometimes it absolutely does not.
How I Would Actually Use AI to Plan a Hike
I don’t think hikers need to ban AI from trip planning. I think we need to put it in the correct place in the research hierarchy. My process would look something like this:
- Start with the agency responsible for the land. Check the Forest Service, National Park Service, BLM, state park, or other land manager for closures, regulations, permits, access information, and official trail details.
- Find the best local conditions source. On Mount Shasta, that’s the Mount Shasta Avalanche Center and ranger station. Elsewhere, it might be an avalanche forecast center, park conditions page, local trail organization, or ranger district.
- Use a real navigation tool and download the map offline. I use hiking apps too, but I want the route available when cell service disappears. For more consequential backcountry trips, redundancy matters.
- Look for recent firsthand information. Recent trip reports can fill gaps between official updates, particularly for things such as blowdowns, creek crossings, snow patches, or trailhead road conditions. I still treat those as individual reports rather than gospel.
- Then use AI to organize what you’ve learned. Have it build a checklist, identify gaps in your plan, compare information, or suggest questions you should verify. If it gives you a factual answer that matters to your safety, follow the answer back to the original source.
That last part is key. AI can be the assistant. It shouldn’t automatically become the expert. Educate yourself on the trail and terrain, and look to local resources for current information.
Your Phone Shouldn’t Be Your Only Plan Either
The Mount Shasta rescue isn’t solely an AI cautionary tale. It’s also another reminder that outsourcing all of our outdoor competence to a phone creates its own problems. According to reporting on the rescue, the group was using a phone for navigation and encountered battery problems during the descent.
The Mount Shasta Trail Association tells hikers to download or print maps before leaving because cell service can be poor, while National Park Service hiking guidance recommends carrying appropriate navigation tools and knowing how to use them.
The Mount Shasta Avalanche Center’s Ten Essentials list goes even further: map, compass, optional GPS, insulation, illumination, extra food, water treatment, and emergency shelter are all part of basic preparation. It specifically recommends an extra day’s supply of food.
That looks quite a bit different from asking a chatbot for the minimum amount of food you can get away with carrying.
Don’t rely on a single source of information, either. Always have a secondary source for navigation.
AI Doesn’t Know When You’re In Over Your Head
This may be the bigger lesson.
Experienced hikers aren’t safe because they possess a perfect packing list. They’re safer because experience teaches them when something doesn’t feel right.
Weather is moving in faster than expected. The creek is higher than it should be. You’re covering fewer miles per hour than your itinerary assumed. That snowfield looks considerably steeper in person. You’re more exhausted than you expected to be. The route has disappeared. Or it’s 1 p.m., you’re still 3,000 vertical feet below a 14,179-foot summit, and the responsible decision is to turn around.
An AI-generated itinerary can’t make that judgment for you. A map can’t make it either.
That’s part of what the Mount Shasta climbers themselves acknowledged afterward when they reportedly told rescuers, “We relied too much on AI rather than our own critical thinking.” That’s probably the most useful line to come out of the whole incident.
So, Should Hikers Use ChatGPT, Gemini, or Other AI Tools?
You can use AI to discover trails. Ask it what questions you should research. Have it organize your packing list. Ask it to explain an unfamiliar piece of gear. Give it information you’ve already verified and have it turn that information into a trip plan.
Just keep the relationship straight.
When ChatGPT tells me what rain jacket features I should consider, that’s useful.
When it tells me that the Mount Shasta Avalanche Center says a particular route is currently unsafe, I’m opening the Avalanche Center’s website and reading what the Avalanche Center actually says.
And if the AI-generated infographic tells me to climb one route while the text underneath tells me not to climb that route?
I’m definitely not heading for the trailhead.
AI can help you organize the reliable information you collect, but it should not be the number one source of the information. It cannot be trusted to be 100 percent accurate and up-to-date.
FAQ: Using AI for Hiking and Backpacking
Can I use ChatGPT or Gemini to plan a hike?
Yes, but AI is best used as a planning assistant rather than an authoritative source. It can help you find topics to research, organize an itinerary, create a preliminary packing list, or suggest questions to ask. Verify safety-critical information such as route conditions, closures, weather, permits, water sources, and required equipment with official or trusted local sources.
Is AI reliable for current trail conditions?
Not reliably enough to use as your only source. Trail conditions can change because of weather, snow, wildfire, flooding, maintenance, road closures, or other factors. Even an AI system with internet access can misinterpret or combine information from different dates. Check the relevant land manager, ranger district, avalanche center, or local trail organization directly.
Can AI tell me how much food and water to bring hiking?
AI can help you think through food and hydration planning, but you shouldn’t blindly follow a generated quantity. Water and calorie needs depend on the person, weather, altitude, duration, exertion, available water sources, and the possibility that a trip takes longer than planned. For remote trips, plan for unexpected delays rather than only the best-case itinerary.
Is AllTrails enough for backcountry navigation?
Hiking apps are useful, but no single phone app should be your entire navigation strategy on consequential backcountry trips. Download maps for offline use, keep your phone charged, and consider additional navigation methods appropriate to the terrain. Most importantly, know how to use the tools you carry.
What sources should I use when planning a hike?
Start with the agency managing the land, such as the Forest Service, National Park Service, BLM, or state park agency. For specialized terrain, also look for local avalanche centers, ranger stations, trail associations, current weather forecasts, and recent condition reports. For Mount Shasta specifically, the Mount Shasta Avalanche Center provides current climbing conditions and route information.
Can AI-generated hiking maps be wrong?
Yes. AI can create convincing-looking maps, graphics, and itineraries that contain outdated or contradictory information. In my own test, ChatGPT generated a Mount Shasta infographic using Avalanche Gulch even though the accompanying text correctly warned that Avalanche Gulch was not recommended under the current conditions. Treat generated graphics as illustrations unless you’ve independently verified the information they contain.
What are the Ten Essentials for hiking?
The Ten Essentials generally cover navigation, sun protection, insulation, illumination, first aid, fire, repair tools, nutrition, hydration, and emergency shelter. The exact equipment you need depends on the outing, but the system is intended to help hikers prepare for unexpected weather, injuries, delays, and emergencies.
