Nobody Mentioned the Snow.
ChatGPT and Google Maps chose the route up Unnecessary Mountain. Neither said anything about spring snow at elevation, so the pair went up in flat-soled sneakers.
What happened on Unnecessary Mountain
Unnecessary Mountain sits in the Coast Mountains above Lions Bay, north of Vancouver. In spring 2025, two hikers used ChatGPT together with Google Maps to choose a route up it. Neither tool told them there would still be snow at the top, and they set off in flat-soled sneakers.
They got into terrain they couldn't handle in that footwear and called for help. Lions Bay Search and Rescue went up to them — carrying, among other things, the right boots.
Brent Calkin, who leads the team, told the Vancouver Sun: “We ended up going up there with boots for them. We asked them their boot size and brought up boots and ski poles.”
In May 2025 the team issued a public warning about exactly this pattern — hikers arriving on mountains with routes chosen by AI tools and phone maps that have no way to account for snowpack, terrain difficulty or changing weather.
What Calkin actually said about AI
His broader point isn't that the tools are useless. It's that the volume of available information makes it very easy to get out of your depth: “With the amount of information available online, it's really easy for people to get in way over their heads, very quickly.”
When he tested ChatGPT himself, he found that the quality of the answer depended heavily on the quality of the question — which is a problem, because knowing what to ask is precisely the expertise a beginner doesn't have yet. (That framing is the reporting's summary of his testing, not a direct quotation; the quoted lines on this page are his own words.)
What he contrasts it with is worth reading twice — the value of asking an actual person:
Someone might tell you there's a storm coming in this week. Or I was just up there Wednesday and it looks good. Or you're out of your mind, don't take your six-year-old on that trail.
What a chatbot can't know about a spring snowline
None of this is exotic information. It's published, current, and local — three properties a general assistant answering from training data cannot have.
That Winter Is Still Up There
Snow lingers at elevation for months after it's gone from the trailhead. Where the snowline sits this week is a real, measurable, locally-published number — and it is not something a model can recall, because it changes every week.
That a Road Map Isn't a Trail Map
Google Maps was built for streets. Pointed at a mountain it will happily return a line and a time estimate, neither of which accounts for snowpack, scrambling, or the difference between distance and effort.
What's on Your Feet
The gap between sneakers and boots decides whether a snow slope is a nuisance or a rescue. Nothing in a chat exchange looks at your feet, and a plan that never raises footwear reads as though footwear doesn't matter.
What Somebody Saw Up There on Wednesday
The most valuable trail information is days old and comes from a person who was just there. It lives in conversations and local reports, not in training data, and it is the single thing most likely to change your plan.
The question you didn't know to ask
Calkin's finding — that a good answer depends on a good question — is the whole problem stated cleanly. A chatbot answers what you asked. “What's a good hike near Vancouver this weekend?” gets you a good hike. It does not get you “there's still two feet of snow above 1,200 metres and you'll need traction,” because you didn't ask, and nothing in the exchange volunteers it.
Experienced hikers ask about the snowline without thinking, the way they check the wind. That habit isthe expertise. A beginner who hasn't got it yet will get a plausible answer to a slightly-wrong question and no signal that anything is missing.
Which is why a tool aimed at the outdoors should be built to raise the thing you didn't ask about — and to say clearly when it has no data on something that matters.
Already have a plan a chatbot gave you?
Paste it into Roamze's plan check and we'll check what we can against the real forecast, avalanche and snowpack data for your route and dates — including the snowpack numbers that decide whether you need traction — and tell you plainly what we couldn't check.
It is not a substitute for your own judgment, and it is not a substitute for the conversation Calkin is describing. Call the local SAR group or ranger district. They know what this week looks like.
Before a spring or early-season hike
- Ask where the snowline is right now, not what the trail is like in August
- Use a mapping app built for trails — Lions Bay SAR points people to AllTrails, Gaia GPS and Avenza rather than a road-navigation app
- Carry traction and real boots if there's any chance of snow at elevation
- Find one recent trip report from someone who was actually up there
- Turning back because the snow started earlier than expected is a normal outcome, not a failed trip
Sources
Other AI planning failures
Same pattern, different mountains — a confident plan from a tool with no way to check it.
Mount Shasta, California
Gemini planned the climb and under-provisioned food and water. Three hikers spent the night stranded in Mud Creek Canyon.
Read guideTatra Mountains, Poland
A chatbot's shortcut to the Valley of the Five Lakes put two hikers on a traverse they couldn't finish. Helicopter evacuation.
Read guideTonggongjian, China
A chatbot matched a 1,500m peak to a man with no hiking experience, and never mentioned gear. He arrived in skate shoes.
Read guideWhat Is an AI Trip Planner?
How these tools work, and what separates one that helps from one that just sounds confident.
Read guideHave a Chatbot Plan? Get a Second Opinion.
Paste the itinerary an AI gave you. Roamze checks what it can against real forecast, avalanche and snowpack data for your dates — and tells you what it couldn't check.