Researchers have nearly doubled the known count of figurative geoglyphs etched into Peru’s Nazca desert, bringing the total to roughly 800 giant figures that span an arid plateau south of Lima. The findings, published in the Proceedings of the National Academy of Sciences, were driven by an AI-assisted survey that spotted dozens of previously unrecognized shapes invisible to the naked eye on the ground. Despite the new data, no one has produced a definitive explanation for why ancient people spent centuries scratching enormous animals, plants, and abstract forms into the earth.
An AI survey rewrites the scale of the Nazca puzzle
For decades, the official inventory of figurative Nazca geoglyphs hovered in the low hundreds. That number changed sharply when a team led by researchers at Yamagata University trained machine-learning models on high-resolution aerial imagery and then verified candidates through fieldwork. Their AI-accelerated survey nearly doubled the number of known figurative geoglyphs, according to the university’s institutional release. The jump matters because it shifts the statistical picture of where figures appear, how large they are, and what subjects they depict.
Many of the newly identified figures are small, sometimes only a few meters across, and sit along ancient footpaths rather than on the broad, flat pampa where the most famous lines stretch for hundreds of meters. That spatial pattern feeds a growing argument among specialists: the geoglyphs may have served different audiences and different purposes depending on their size and placement. Large trapezoidal clearings visible from hilltops could have addressed communal gatherings, while smaller animal and human figures along trails seem oriented toward individuals or small groups walking past them.
How the desert preserved two thousand years of markings
The physical mechanism behind the Nazca Lines is straightforward. According to NASA Earth Observatory, the geoglyphs were made by removing dark surface stones and soil to expose a lighter substrate beneath. The surrounding desert receives almost no rain, so the contrast between the cleared paths and the untouched ground has survived for roughly two millennia. Most scholars date the bulk of the work to the Nazca culture, active between approximately 200 BCE and 500 CE, though some simpler lines may be older.
That preservation is a double-edged research asset. The lines remain visible, but the same dry conditions that protect them also limit the organic material available for radiocarbon dating. No publicly available excavation logs from the latest Yamagata-led survey detail soil samples or artifact recovery at the newly mapped sites. Without that granular data, researchers can describe what was drawn and where it sits but struggle to pin individual figures to specific decades or generations.
Water routes, not star charts, as a working explanation
The stage-one hypothesis tested by the new mapping data is that the figures cluster along seasonal drainages in patterns aligned with ancient water-flow models, suggesting they functioned as route markers for small ritual processions tied to water availability rather than astronomical calendars. Earlier theories linking the lines to solstice alignments or constellation maps have not held up under statistical scrutiny; only a handful of lines point toward significant celestial events, no more than random chance would predict.
The Yamagata team’s institutional release states that the new figures suggest the lines were tied to small-group rituals rather than large public events. If the newly mapped geoglyphs do track seasonal watercourses, the implication is practical: people walking between water sources during dry months would have encountered figure after figure, each one reinforcing a shared cultural narrative about the land they crossed. The lines, in this reading, were not meant to be seen from the sky at all. They were meant to be walked.
No single dataset has closed the case. The water-route model fits the spatial distribution better than astronomical alternatives, but it still rests on correlation between geoglyph locations and reconstructed drainage patterns rather than on direct textual or material evidence of intent. The Nazca left no written language, and the few ceramic and textile motifs that echo geoglyph subjects do not come with captions explaining why a spider or a hummingbird earned a place on the desert floor.
Gaps that 800 figures still cannot fill
Three significant holes remain in the evidence. First, no official Peruvian government archaeological registry entries or permit records have been published that would verify the exact coordinates and condition assessments of all 800 figures. Second, calibrated radiocarbon dates tied to specific geoglyph construction events are absent from recently released datasets, meaning the broad 200 BCE to 500 CE window cannot yet be narrowed for individual figures. Third, the AI model that detected new geoglyphs was trained on shapes already known, which introduces a selection bias: it may be better at finding figures that resemble existing ones than at spotting entirely new categories of markings.
The Yamagata team’s PNAS paper represents the largest single expansion of the Nazca catalog, but its analytical reach stops at distribution patterns and subject classification. The deeper question of motive, why a culture invested generations of labor in figures most members could never see whole, has no answer grounded in direct evidence. Competing hypotheses range from fertility rites to territorial markers, and none can be decisively confirmed or rejected with the material now in hand.
AI’s promise and limits in the desert
From a methods perspective, the Nazca project shows how machine learning can accelerate archaeological discovery without replacing traditional fieldwork. The algorithm flagged candidate shapes, but human teams still had to hike out, confirm that the marks were cultural rather than natural, and record their condition. In that sense, the desert plateau has become a test bed for how AI might be used to scan other arid landscapes for traces of past societies.
Yet the same tools that make it easier to find more lines also complicate interpretation. When models are trained primarily on known motifs-hummingbirds, monkeys, fish, and stylized humans-they may systematically overlook idiosyncratic or experimental figures that fall outside those patterns. That could skew our picture of Nazca creativity toward the canonical and away from the unusual. Future surveys will likely need to vary training data and incorporate unsupervised techniques to minimize that bias.
Preservation pressures on a fragile archive
Even as digital tools multiply the known geoglyphs, the physical lines remain vulnerable. Vehicle tracks, informal tourism, and nearby development all threaten to scar or erase the shallow grooves that define the figures. International agencies and Peruvian authorities have periodically turned to satellite monitoring and other space-based resources to track new disturbances, but enforcement on the ground is uneven across such a wide and sparsely populated area.
Remote-sensing specialists within the broader Earth science community have argued that regular high-resolution imaging could function as an early warning system, flagging tire marks or unauthorized excavations before they become permanent scars. That approach requires sustained funding and coordination, as well as agreements on how frequently to image the region and how to share alerts with local heritage managers.
For now, the Nazca Lines occupy an uneasy position between scientific dataset and cultural symbol. They are at once a laboratory for AI-assisted archaeology, a fragile archive of ancient ritual life, and a magnet for speculative theories that outpace the evidence. The new tally of roughly 800 figurative geoglyphs sharpens the mystery rather than resolving it, revealing a denser and more varied landscape of images than researchers had previously mapped. Until excavations, dating campaigns, and careful on-the-ground documentation catch up with what algorithms can see from above, the desert drawings will continue to invite more questions than they can answer.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
