AI GUIDEPartner content

AI Agents for Finding Archaeological Sites in LiDAR: Methodology and Limits of Verification

AI agents can speed up the initial search for possible structures in LiDAR data, but they do not confirm an archaeological find. An analysis of an experiment involving roughly 1 200 km², 48 agents, and a previously published site shows why coordinates, academic literature, and hypotheses need to be checked with archaeologists.

Affiliate link: your price stays the same and the project earns a commission.

Finding archaeological sites in LiDAR data with AI agents is used for the preliminary analysis of remote-sensing data and the selection of areas for further investigation. It does not confirm that an ancient structure is actually present on the ground, and it does not replace archaeological research.

This is preliminary screening, not confirmation of a find.

How LiDAR data searches work

LiDAR measures the distance to a surface using laser pulses. When data are processed to reduce the effect of vegetation, terrain images may reveal features hidden beneath the forest canopy. In archaeological searches, it is important to consider not only the images themselves, but also their coordinates, resolution, coverage, processing method, and information about which areas were actually examined.

In the research experiment described, developer Pietro Miotti used open NASA data and AI agents to analyze an area of Maya territory covering roughly 1 200 km². According to his account, he turned to this area after failing to find enough accessible data for Peru. He used 48 agents to search for possible structures and did not use a specialized archaeological model.

The number of agents alone does not indicate the accuracy of the result. It describes the configuration of a specific experiment, while search quality depends on the input data, the criteria used, terrain processing, and how carefully people check the candidates identified.

What 48 AI agents can do

In this case, the agents searched for signs of possible pyramids, plazas, platforms, and other structures, while also ruling out features that might be explained by natural terrain. This approach distributes the initial review of a large dataset across many tasks: individual agents assess areas or features, after which the researcher compiles and checks the results.

The result of this work is not a completed archaeological conclusion, but a list of hypotheses linked to coordinates. For the hypotheses to be verifiable, the report should show the original image section, the processed image, the location of the suspected feature, and the reasons the algorithm identified that form. If an agent provides only a label such as “pyramid” without retaining the data behind it, an expert cannot reproduce the reasoning.

What features the agents identified

According to Miotti, the agents pointed to a possible settlement on a hilltop. The hypothesis included a flat platform, an oval structure, a raised inner courtyard, and a depression that may have been used to store water. The agents estimated that the suspected pyramid was at least 8–12 m high. This is an estimate based on data analysis, not the result of a field measurement.

The agents then proposed historical interpretations: they assigned the settlement to the Classic Maya period, approximately 250–900 CE, described the pyramid as a possible temple, the plaza as a place for ceremonies or trade, and the courtyard as the residence of a local ruler. These remain hypotheses proposed by the experiment’s participants. The shape of the terrain alone does not establish a structure’s purpose, age, or connection to a particular political organization.

The agents also reported that the structures were beneath forest 10–17 m tall. This claim describes the presumed conditions at the site, but does not independently verify the presence of the features or confirm the archaeological interpretation. Assessment requires the data themselves, details of the processing method, and expert comparison with materials concerning the area.

Why review by an archaeologist is necessary

A feature in LiDAR data may correspond to a structure, but it could also be explained by natural terrain, later earthworks, or data-processing artifacts. An archaeologist compares suspected structures with local history, known sites, characteristic settlement forms, and available survey materials. If needed, the specialist assesses what further verification methods are appropriate and necessary.

An AI agent does not confirm a find.

The algorithm identifies a candidate; specialists check the scientific interpretation against sources and field data. The absence of a site from an official register also does not prove that it has not been described in academic literature or marked on a research map.

In the case described, Miotti sent the materials to the Mexican authorities on 20 October. The next day, Mexico’s National Institute of Anthropology and History replied that its archives contained no registered archaeological sites at the specified coordinates. The response concerned records in the institute’s archives, not a comprehensive review of publications and research maps.

How to check the literature and coordinates

The experiment’s history shows why searches of official registers and academic literature need to be conducted separately. On 11 November, it emerged that archaeologists had already mapped the site in a study by Belli and colleagues, published in 2025. The agents had missed this work in their literature search. The result was not the discovery of an unknown settlement: according to the experiment’s account, the system independently reproduced a previously published finding.

Before declaring a site new, a researcher should check both its appearance and its coordinates. An incorrect coordinate system or imprecise comparison of areas can lead to a mistaken conclusion that the terrain feature identified matches a known site. Results should be compared with academic publications, maps, and registers, and any differences in how coordinates are recorded should be clearly documented.

  • Check publications. Search for studies on the region, nearby sites, local place names, and the remote-sensing data used. Account for variations in how place names and site names are spelled.
  • Verify the coordinates. Record the original coordinate system, the transformation method, and the point or boundary of the area analyzed. Do not treat an approximate match as sufficient confirmation.
  • Compare sources. Check available research maps and official records separately: the absence of an entry in one register does not mean there are no publications.
  • Keep the basis for the hypothesis. Retain the original materials, processed images, settings, and agent outputs so specialists can assess the basis for the interpretation.
  • Separate observation from explanation. “The image shows a flat platform” describes a feature; “this is a ceremonial center” is already a historical explanation that requires further evidence.

What the result of this experiment means

Miotti says he considers the experiment successful because, without archaeological training, he organized the analysis of a large dataset and received a response from a government institution. This is his assessment of the process, not proof that the system discovered a new site. The publication about the same location that was identified later demonstrates a specific limitation: the agents helped formulate and describe a hypothesis, but did not conduct a sufficiently comprehensive literature review.

For archaeologists and geospatial researchers, the practical value of this approach lies in the initial selection of areas, not in automatically classifying them as archaeological sites. If candidates are accompanied by coordinates and verifiable materials, agents can make it easier for experts to review a large area. Without checks of publications, original images, and archaeological interpretation, however, a convincingly worded report risks presenting a known site as a new discovery.

Frequently asked questions

Can AI agents independently confirm an archaeological find?

No. They can identify characteristic forms in the data and propose hypotheses. Specialists determine archaeological status and interpretation by comparing images with academic information and survey results.

Why check publications if an official register contains no record?

Registers and academic literature serve different purposes and may contain different information. In the experiment described, the absence of a registered site in the Mexican institute’s archives did not bring to light a 2025 study in which the site had already been mapped.

What should be published alongside search results?

At a minimum, a useful report links each hypothesis to coordinates, source data, processed images, and an explanation of the features identified. It should also state which publications and maps were checked and which conclusions remain tentative.

Compare models before you start

The service sets its plans, limits and model catalog. If they differ from this article, contact us so we can update it and record a new review date.

Browse models

Affiliate link: your price stays the same and the project earns a commission.

AI agents LiDAR data archaeology archaeological sites remote sensing verification of archaeological hypotheses NASA data

SEO Mind42 editorial team

We explore SEO and neural networks in practice: test services on our own projects, verify prices and limits against primary sources, and share things you can put to use the same day.

📚 Reference guide to SEO and AI 🔄 Materials are updated 🕐 Updated: 21 November 2026

Related reading

All in this section →