Google researcher uses 48 AI agents to flag possible unmapped Maya site in Mexico

From a vacation thought to an archaeology lead

A Google researcher says a swarm of AI agents helped him identify what may be an unmapped Maya site in Mexico, according to TechRadar. Pietro Miotti described the project as starting from a random thought while on vacation, in which he wondered whether AI agents could be set loose on LiDAR datasets.

LiDAR, or Light Detection and Ranging, uses laser pulses to map terrain and can reveal shapes hidden beneath cloud cover or forest canopy, including features that are hard for archaeologists to reach on foot. Miotti's experiment suggests such data can be screened at speed by software agents before experts ever visit a site.

How the agent search worked

Miotti assigned 48 agents drawn from different AI apps to comb through a NASA database of LiDAR information collected across Peru, TechRadar reports. The agents returned several hits on structures that could be buildings, and a second pass narrowed the search to one particular site that has not yet appeared in scientific literature.

The agents did more than scan. According to TechRadar, they identified Mexico's National Institute of Anthropology and History, known as INAH, as the right body to contact, and then drafted an email in Spanish describing the find and its location on the map. The whole sequence, from the initial idea to a reply from INAH, took less than 72 hours, Miotti said in a post cited by TechRadar.

What is confirmed and what is not

The key caveat is that the site has not been confirmed as an ancient Maya site. TechRadar reports that INAH experts are investigating further, and that the location is a promising lead rather than a verified discovery. Readers should treat the Maya identification as a claim under review, not an established fact.

What is documented is the workflow: a researcher used commercial AI agents to triage public LiDAR data, then handed a candidate location to the relevant national authority. The archaeological conclusion will depend on fieldwork and expert assessment by INAH.

A wider signal for AI in research

Miotti framed the result as evidence of a broader shift, telling TechRadar that he believes science is on the brink of a revolution across disciplines, not just coding, physics and math. He also suggested that the entire planet could be mapped with LiDAR and handed to agents to help find what remains.

He went on to use Gemini to generate an image of how the site might have looked around 700 CE, when the Maya were living there, TechRadar reports. That image is an illustration, not evidence, and should be read alongside the LiDAR scan rather than as a reconstruction of a confirmed site.

Why it matters

The story illustrates how AI agents are being used to spot patterns in very large datasets, a task they are well suited to. It also shows a practical division of labour: machines flag candidates at scale, while national heritage bodies such as INAH decide what is real and what happens next.

For readers, the takeaway is cautious optimism. The speed is notable, but the scientific value will be settled by experts on the ground, not by the agents that raised the flag.