AI Found 303 Nazca Figures in Six Months. The Last Century Found 430.

The number that made the headlines

Since the Nazca lines of southern Peru entered the scientific record about a century ago, careful survey had identified on the order of 430 figurative geoglyphs, the ones that depict something: a bird, a monkey, a killer whale, a human figure. In a study published in the Proceedings of the National Academy of Sciences in September 2024, a team from Yamagata University's Institute of Nazca and IBM Research reported 303 more, confirmed by field survey in about six months.

That is roughly a doubling of the known figurative geoglyphs, accomplished at something like sixteen times the previous rate of discovery. It is a genuinely large result, and it is easy to tell it wrong.

What the AI actually did

The machine did not walk the desert and announce a monkey. What the team built was a detector: a model trained on the geoglyphs already known, then turned loose on high-resolution aerial imagery of the Nazca Pampa to flag candidates, faint marks with the right shape and scale that a human might spend a lifetime never noticing. The AI's job was to shrink an impossibly large search space down to a list of places worth visiting.

Then people went and looked. Every one of the 303 was verified on the ground by archaeologists, because a suggestive smudge in a photograph is a hypothesis, not a discovery. The right way to describe this is not "AI discovered 303 geoglyphs." It is that AI proposed and humans confirmed, a division of labour, not a replacement of one by the other. Framed that way, it is still remarkable: the bottleneck in Nazca research was never the walking, it was knowing where to walk.

The finding underneath the finding

The count is the headline; the interesting part is what the new figures are like. The geoglyphs the survey added were overwhelmingly the smaller relief-type, made by clearing stones from a patch of ground, typically only a few metres across, and clustered along the winding trails that people once walked on foot. Their subjects skew human and domestic: people, llamas, and scenes of everyday and ritual life, at a scale meant to be taken in up close, by an individual or a small group.

That contrasts with the enormous line-type geoglyphs, the famous ones, drawn along the network of dead-straight lines, big enough to read only from the air or the surrounding hills, and far more likely to depict wild animals. The pattern the study draws out is that the two kinds may have done two different jobs: the giant line-figures as communal markers along ceremonial pathways, the small relief-figures as something more personal, encountered on the move.

A map of where the small figures cluster is, in effect, a map of where people walked.

Why it belongs in the record, carefully

None of this settles the old, tired question of "who the Nazca lines were for", a question that has attracted more pseudoscience than almost any site on Earth. The people who made them were the Nazca culture, roughly 2,000 years ago; there is no mystery of authorship to solve, only detail to recover. What the 2024 work adds is detail, at a rate that suggests the catalogue is still far from complete. The researchers are explicit that many more likely remain, below the threshold their imagery and model could resolve.

The cautious reading is the accurate one. AI did not rewrite Nazca; it accelerated a survey that people still had to finish with their boots. But "accelerated a century of work into six months" is not a small sentence, and it is the true one.