The project and the claim
According to The Verge, Google researcher Pietro Miotti said he put 48 AI agents to work analyzing NASA lidar data in search of undocumented archaeological sites in the Maya region, after an initial interest in Peru, which is less covered by public lidar data. The agents reportedly flagged several promising locations, including an alleged village not present in the registries. Miotti specified that it was a personal project, without the use of Google resources or computing capacity, and that he contacted Mexico's National Institute of Anthropology and History (INAH) to verify whether the site was already known, receiving a response that no site was registered at that location.
The researcher stressed that he is not an archaeologist and that the find would have required field verification, while still presenting the result as evidence of the potential of AI agents to accelerate scientific exploration. The story, however, took a turn when it emerged that the site had already been documented in an academic journal more than a year earlier.
The archaeologists' criticism
Several voices in the archaeological world criticized the episode on social media, as The Verge reports. Sarah Parcak, professor of archaeology at the University of Alabama at Birmingham, recalled the importance of comparing results with existing research and the role of peer review, observing that the absence of a site from the Mexican government database does not imply that it was unknown or unmapped. Other archaeologists spoke of misinformation and undue credit to those who have used advanced digital tools for decades.
Miotti replied to the publication that he had updated the post and the blog to acknowledge the archaeologists' work, explaining that, as a non-specialist, he had not understood the difference between an unregistered site and an unmapped site. He also pointed to a technical limitation of his tools: because archaeologists do not publish the coordinates of sites, verification requires comparing images with figures from scientific articles, a task that is more difficult for the agents. Elizabeth Graham, professor of Mesoamerican archaeology at University College London, observed that the problem is not discovering new sites, because many scholars know numerous unexplored ones, and criticized the use of the term discovery, recalling that local communities often already know the sites in their own territory. She also acknowledged that Miotti had done decent work mapping mounds from lidar data, while stressing that the real obstacle is the lack of funding for excavations and mapping.
The verification issue and the risks
Ivan Šprajc, professor of archaeology at the Research Centre of the Slovenian Academy of Sciences and Arts and co-author of the article that documented the site, told The Verge that he was not particularly impressed by the agents' work, recalling that archaeologists have long used lidar and other remote sensing tools and that merely identifying something does not add much to knowledge. While calling the check with INAH commendable, he said it is not sufficient, because the institute does not have all possible information. Šprajc expressed concern about the risk that growing use of AI tools by amateurs could inundate archaeologists with reports, citing that he had received that very day an email from another alleged AI discoverer, and recalled the danger that publicity about archaeological sites attracts looters.
The case, according to The Verge, reflects a broader tension in science: increasingly powerful AI tools allow non-specialists to produce even notable results, but without comparison with existing literature and without the context that an expert can provide. It remains to be verified, on the merits, what the actual scope of the mapping work was and how the AI agent workflow can be integrated into scientific validation procedures, a topic that the material raises but does not resolve.
What remains open
The news story presents a case with some elements confirmed and others still to be clarified. It is confirmed, according to The Verge's account, that the site was already documented in an academic publication and that archaeologists criticized the claim. The statements about the personal nature of the project and the absence of Google resources are the researcher's. The question remains open, however, of how AI agents can be reliably used to verify existing knowledge, without replacing specialist expertise or fueling expectations of discoveries that do not add knowledge.
