A Maya 'discovery' made with 48 AI agents had already been published in 2025

What Miotti claimed

As infosecu.technews.tw reported, citing The Verge, Google researcher Pietro Miotti said this week that a team of AI agents had found a possibly unmapped Maya site in Mexico. He stressed it was a personal hobby project and used no Google computing or other resources.

The idea came to him while travelling in Peru. He wondered what historic explorers would do with modern tools. He chose LiDAR, a laser technique that builds detailed 3D maps of the ground. Public LiDAR coverage in Peru was too thin, so he turned to the Maya region.

How the check went wrong

The 48 agents combed NASA's LiDAR data and flagged several places, including a large settlement that looked unregistered. Miotti then had the agents write a letter in Spanish to Mexico's national anthropology and history institute, INAH. It replied that it had no record of a registered site at that location.

He wrote that the process took under 72 hours. But the site had already been described in a 2025 academic paper, and archaeologists were quick to say so.

What archaeologists said

Sarah Parcak of the University of Alabama at Birmingham said the episode shows why peer review exists. She added that a site missing from a Mexican government database is not necessarily unknown or unmapped. Archaeologists usually keep site coordinates private, so confirming novelty means comparing images against figures in papers, which is much harder for an AI agent than querying a database.

Elizabeth Graham of University College London questioned the word "discovery", since local communities often know these places already. She said funding is a bigger obstacle than finding more sites. Ivan Šprajc, a co-author of the 2025 paper, said archaeologists have used remote sensing for some time. He fears being buried under amateur claims, and he received another from an AI "discoverer" the same day. He also noted that publicizing locations can invite looting.

Miotti's response

Miotti acknowledges that he is not an archaeologist and that any find would need fieldwork to confirm. He said he did not know "unregistered" and "unmapped" were different things, and told The Verge his agents and he both missed it.

He has updated his post and blog to credit the earlier work. He still argues that science is near a revolution and wants LiDAR mapping of the whole Earth, with AI agents helping to find the rest.