The vast desert etchings known as the Nazca Lines, some of the most famous archaeological features on Earth, have yielded a wave of previously unknown figures after researchers trained artificial intelligence to hunt for them in aerial imagery. Working across the arid plains of southern Peru, a team nearly doubled the known catalog of figurative geoglyphs in a matter of months, spotting shapes that had eluded detection for centuries. The effort demonstrates how machine learning can accelerate discovery in a landscape where subtle, faded markings sprawl across an area far too large for humans to survey by eye alone.
What the geoglyphs are
The Nazca Lines are enormous designs etched into the desert floor more than 2,000 years ago by the pre-Inca Nazca culture, created by removing the dark, oxidized surface stones to expose the lighter ground beneath. They range from long straight lines and geometric shapes to figurative images of animals, plants, and human-like forms, some stretching hundreds of feet across. Their scale is so great that many are best appreciated from the air, and their purpose remains debated, with theories tied to water, ritual, and astronomy. The site’s cultural importance is recognized by UNESCO, which lists the Lines and Geoglyphs of Nazca and Palpa as a World Heritage property.
How AI found the new figures
The recent discoveries came from a collaboration that paired archaeologists from Yamagata University’s Nazca research effort with computing specialists who built and trained a deep-learning model to recognize the faint signatures of geoglyphs in high-resolution aerial photographs. Rather than replacing fieldwork, the system acted as a filter, flagging candidate locations across a huge volume of imagery so that researchers could concentrate their on-the-ground verification on the most promising spots. The technical collaboration involved IBM Research, whose scientists worked with the archaeological team to teach the model what a geoglyph looks like from above despite erosion and centuries of accumulated debris.
The payoff was substantial: the AI-assisted survey identified hundreds of previously unrecorded figures in roughly half a year, effectively doubling the count of known figurative geoglyphs. The newly spotted designs included depictions of humanoid figures and animals, some smaller and more weathered than the celebrated large-scale images that first made the site famous, and therefore easy to overlook without computational help.
Why the technique works here
The Nazca plateau is an ideal proving ground for this kind of analysis. The geoglyphs are numerous, spread across a wide and difficult terrain, and often so degraded that they blend into the surrounding desert, exactly the conditions under which a trained model can outperform manual searching. Documentation of the computing effort noted that the deep-learning training ran on high-performance hardware, and an account published by NVIDIA described how the model was taught to distinguish genuine markings from natural features and modern disturbances. By narrowing millions of image tiles down to a manageable set of leads, the system compressed what would have been years of human survey into a far shorter campaign.
What the discoveries add
Beyond the raw increase in numbers, the newly found figures give archaeologists more material to study the meaning and distribution of the geoglyphs. Researchers have observed that many of the smaller figurative designs cluster near winding paths, suggesting they may have been made to be viewed by people traveling on foot rather than only from a distance, a clue to how the Nazca people may have used and encountered them. Reporting by France 24 conveyed how the discovery reframed the scale of the site, showing it to be even richer than the famous large animal figures had suggested.
The work also points toward a broader future for the field, in which AI-assisted analysis of aerial and satellite imagery helps locate archaeological features across other landscapes. For Nazca specifically, the combination of machine learning and careful ground verification has turned a well-studied wonder into a site still actively giving up new secrets, with researchers expecting that further surveys could reveal still more figures hidden in the desert.
Machine learning as an archaeological tool
The Nazca survey has become a frequently cited example of how artificial intelligence is reshaping archaeology, a discipline traditionally bound by the slow pace of manual survey and excavation. Trained on examples of known geoglyphs, a model can sift through enormous volumes of aerial and satellite imagery far faster than any team of researchers, surfacing candidates that human eyes would likely miss amid erosion and desert clutter. The approach does not replace the archaeologist, whose judgment remains essential to confirm a find and interpret its meaning, but it dramatically narrows where that expert attention should be directed.
Researchers see broad potential in the method beyond Peru, from mapping buried structures to detecting looting damage and monitoring fragile heritage sites over time. For Nazca specifically, the collaboration has reframed a site studied for nearly a century as one still capable of surprising researchers, with the desert expected to reveal additional figures as surveys continue and the models improve.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- Card skimmers hidden on gas pumps and ATMs are draining accounts, and here’s the tell
- The FBI says hackers are hijacking outdated home routers, and it named the models to check
- Older Teslas are wearing out in ways early owners never saw coming
- A common childhood virus is now tied to multiple sclerosis years later
