The AI Archaeologist: How Neural Networks are Uncovering Lost Cities from Satellite Data

For centuries, the discovery of a lost city or an ancient settlement was the stuff of legend—the result of a lifetime of painstaking research, grueling expeditions, and a healthy dose of luck. Archaeologists would spend years trekking through dense jungles or scanning vast deserts, hoping to stumble upon a clue.

Today, some of the most significant archaeological discoveries are being made not with a shovel and a brush, but with a powerful new kind of explorer: an Artificial Intelligence.

By training sophisticated neural networks to analyze immense datasets of satellite imagery and LiDAR scans, a new field of “AI archaeology” is emerging. These digital archaeologists can scan an entire country in a matter of hours, finding subtle patterns invisible to the human eye that point to the ruins of civilizations long lost to time.

The Data: Eyes in the Sky

The first ingredient is the data. Researchers are now using high-resolution imagery from a variety of sources:

• Satellite Photography: Decades of archived satellite photos from agencies like NASA and commercial providers offer a bird’s-eye view of the planet.
• LiDAR (Light Detection and Ranging): This is the game-changer. LiDAR can effectively “see through” dense forest canopies by bouncing laser pulses off the ground. The resulting data creates a detailed 3D map of the topography below, revealing the rectangular outlines of ancient buildings and roads that are completely hidden by jungle.

The Brain: How the AI Finds a Ruin

A human looking at a satellite image of a dense forest might see nothing but trees. An AI, however, can be trained to find what’s hidden.

The process is similar to how facial recognition works. Researchers “train” a neural network by showing it thousands of examples of known archaeological sites. The AI learns to recognize the subtle signatures of human settlement—faint changes in soil color that indicate old earthworks, unnaturally straight lines that suggest ancient roads, or the specific rectangular patterns of buried foundations.

Once trained, this AI archaeologist can be let loose on a massive, unexplored dataset. It can analyze millions of square miles of satellite data, flagging areas with a high probability of containing undiscovered ruins. Human archaeologists can then use these AI-generated maps to target their real-world expeditions with incredible precision.

The Discoveries

This technology is already yielding incredible results. Researchers have used it to identify potential Viking settlements in North America, uncover hundreds of previously unknown Mayan and Amazonian settlements hidden beneath the jungle canopy, and map the full extent of ancient cities that were once only partially understood.

The AI isn’t replacing the human archaeologist. It’s providing them with a superpower. It’s a tool that can process information on a scale and at a speed that was previously unimaginable. By pointing us in the right direction, the AI archaeologist is helping us to piece together the lost chapters of human history, revealing that our ancient world was far more vast and interconnected than we ever knew.