Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
AI helped archaeologists find 303 previously unknown figurative geoglyphs in Peru’s Nazca region during a six-month field survey, nearly doubling the known total. The discovery revealed patterns that support different possible roles for different kinds of figures—but it did not definitively explain why the ancient people made them. The algorithm pointed researchers toward likely sites; archaeologists verified the finds and interpreted what they might mean.
What are the Nazca geoglyphs?
The Nazca (also spelled Nasca) geoglyphs are designs made on the arid landscape of southern Peru. Their makers cleared away dark surface stones to expose lighter ground beneath, creating lines, geometric shapes and figures. Some designs depict animals or people; others are long straight lines and trapezoids. The region’s geoglyphs are part of a UNESCO World Heritage site.
Their creators are not an unknown civilization waiting to be identified. The enduring questions are more specific: why were different designs made, who encountered them, and how did they relate to routes, ceremonies and other features of the landscape? The 2024 research offers evidence about those questions by comparing where geoglyphs occur and what they depict—not by translating their meaning directly.
How AI helped locate the new figures
Yamagata University’s Institute of Nasca and IBM Research used a computer-vision system to analyze aerial imagery and rank places that might contain geoglyphs. The model learned from a limited set of known examples, then searched for similar visual patterns. It was a prioritization tool, not a generative chatbot and not an autonomous archaeologist.
#1 Best Overall
The system identified 1,309 promising candidates. Researchers screened them and conducted field surveys, confirming 303 new figurative geoglyphs in six months. The team reported that researchers had to assess an average of about 36 AI suggestions for each likely candidate. Around one-quarter of the candidates received field-survey attention. That ratio makes the system’s role clear: it narrowed a vast search area to a more manageable set of leads, while human observation established which features were genuine. The team reported a 16-fold increase in discovery rate compared with its earlier approach; that figure describes this project’s reported rate, not a guarantee of how AI performs on other archaeological surveys. Yamagata University’s announcement and the peer-reviewed study in PNAS describe the findings.
In practical terms, the process was: imagery went into the model; the model ranked candidate locations; researchers screened the suggestions; field teams checked sites; and archaeologists analyzed the confirmed examples alongside previously known geoglyphs. A visual match alone cannot establish a feature’s age or cultural meaning.
Rank #2
Two kinds of figures, with different settings
Before this work, the smaller figurative geoglyphs were known in numbers too limited to support strong comparisons. The 303 new examples expanded the dataset enough for researchers to identify differences between two broad categories:
| Type | Patterns reported by researchers | Possible social context |
|---|---|---|
| Line-type | Generally larger; more often depict wild animals; associated with networks of straight lines and trapezoids | The researchers interpret the pattern as likely connected to community-level ritual activity. |
| Relief-type | Generally smaller; more often depict people and domesticated camelids; found near winding trails | The researchers suggest these may have been encountered by individuals or small groups. |
These are tendencies, not rules assigning a single purpose to every image. The location patterns make the interpretation more persuasive than a guess based on one striking figure, but they do not prove exactly what any particular group believed or did. The landscape may have held several overlapping practices rather than one universal function for every geoglyph.
Why the new finds matter—and what they do not prove
Small, faint designs can be difficult to distinguish from natural surface variation, shadows or erosion in aerial photographs. The desert is extensive, and manually inspecting imagery in search of subtle features takes time and sustained attention. Earlier discoveries also shape expectations: a model trained on known forms may be more likely to flag something that resembles those forms than an unusual design that does not.
The expanded inventory matters because it lets archaeologists ask broader questions about distribution and association. It does not, by itself, settle chronology, construction methods, symbolism or ritual use. Those require archaeological judgment and evidence beyond image recognition, including field observation and cultural context.
The candidate-to-discovery ratio also shows why “AI discovered the figures” can be misleading shorthand. Some leads will be false positives; a faint or incomplete figure may be missed, and an unusual image may be rejected because it differs from the examples used to train the system. A natural feature or image artifact can look suggestive from above. Even a correctly located feature could be misclassified by type. Researchers still have to decide what counts as a geoglyph and evaluate what it can support.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Not the first AI-assisted Nazca survey
The 303 discoveries built on earlier work by Yamagata University and IBM. In a feasibility study announced in 2023, deep-learning object detection helped identify four geoglyphs, including a humanoid figure. The team reported that AI-assisted screening was about 21 times faster than manual image analysis by eye in that study. That earlier result established the promise of prioritizing imagery; the later survey applied the approach at a larger scale. Yamagata University’s account links to the earlier work.
More broadly, archaeologists use aerial and satellite imagery, drones, LiDAR and computational methods to map landscapes and identify features for closer examination. Projects such as GeoPACHA show how large-scale imagery survey can combine systematic review with archaeologist-led annotation. Across these applications, software can help researchers search more efficiently; it cannot replace verification or interpretation.
Discovery has a conservation trade-off
Finding more sites can help researchers and authorities document and protect archaeological heritage. But making exact locations widely available can also expose fragile features to vandalism, unauthorized access, looting or excess visitor pressure. Responsible mapping therefore involves more than detection: location data may need to be handled in ways that support conservation without encouraging unsupervised access.
So, did AI solve the Nazca mystery?
Not in the sense of delivering a definitive explanation for the lines. The strongest conclusion is narrower and more useful: AI-assisted searching helped researchers find enough additional figures to reveal a clearer distinction between large designs associated with line networks and smaller figures near trails. The researchers interpret those patterns as evidence that different geoglyphs may have served different audiences and activities. That is a significant step in understanding the Nazca landscape—not a final answer to what every figure meant.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchRead the PNAS study for the full research, or see the open-access paper.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
