Skip to content
Submit incident
AII-1001-37179
ReportedReported by fewer than three independent publishers, with no official finding.

ChatGPT Recommends Non-Existent and Dangerous Hiking Destinations in Peru

ChatGPT recommended tourists visit a non-existent hiking destination in Peru by combining the names of two unrelated locations. Two tourists who followed the advice paid $160 and found themselves stranded in a remote rural area, facing potential danger from high altitude and isolation.

29 September 2025 · OpenAI · Reliability and hallucination
29 September 2025
AII-1001-371792 sources ↓
Quick facts
Category
Reliability and hallucination
Country
Peru
Source
H Kαθημερινή
Date
29 September 2025
Drafted by
Raidu Incident Detection

What happened

ChatGPT generated travel advice recommending a non-existent destination called the 'Sacred Humantay Gorge' in Peru, which the company blended from the names of two unrelated locations.

Two tourists paid approximately 160 dollars based on the ChatGPT recommendation and ended up on a provincial road near Molepata without a guide or real destination.

Reported impact

Miguel Angel Gongora Meza, founder of Evolution Treks Peru, stated that tourists were placed at significant risk because the location does not exist and is in a remote area requiring careful planning. He said the misinformation could have been life-threatening due to high altitude (4,000 meters), lack of oxygen, and lack of cell phone signal in the Andes.

Index analysis

A 2024 study found that 33 percent of users who consulted AI for travel information received inaccurate data. Rayid Ghani, a machine learning professor at Carnegie Mellon University, explained that large language models like ChatGPT generate answers based on statistical word associations without true understanding of the data, leading to hallucinations where plausible but false information is presented.

Classification

Category
Reliability and hallucination
Industry
Travel, leisure and hospitality
Country
Peru
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Economic or property
Harmed
Consumers
Failure modes
Hallucination, Overreliance, Agent misaction, Unsafe recommendation
AI systems
ChatGPT

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

H Kαθημερινή
Σχεδιάζουν ταξίδια μέσω του ChatGPT και καταλήγουν σε μέρη που δεν υπάρχουν | Η ΚΑΘΗΜΕΡΙΝΗ
29 September 2025
newsbomb.gr
Όταν η Τεχνητή Νοημοσύνη σας στέλνει στον Πύργο του Άιφελ... στο Πεκίνο
30 September 2025