Artificial Intelligence as A Game Changer in Archaeology
An excellent example of these difficulties is provided by the Nazca Lines. Discovered in the 1920s, these ancient geoglyphs span hundreds of square kilometres and include both vast line-based designs and smaller relief-style figures. While the larger geoglyphs have been extensively studied, smaller ones often go unnoticed due to erosion, scale, and poor contrast against the desert landscape. Conventional aerial surveys also rely on pilots knowing where to look, making comprehensive mapping difficult.
Recent advances in machine learning and computer vision now provide archaeologists with powerful tools to rapidly process vast datasets with unprecedented accuracy. AI algorithms can analyse high-resolution aerial, drone, and satellite imagery to detect patterns invisible to the human eye.
One groundbreaking case involved Japanese researchers working with IBM to develop an AI model trained to identify faint relief-style geoglyphs—each around 9 meters in diameter—across 600 square kilometres of the Nazca Pampa. These smaller figures are much harder to detect than their line-based counterparts.The AI was trained to recognize faint relief-style geoglyphs, which measure approximately 9 meters in diameter and being much smaller than line-style figures, are considerably more difficult to detect.
Major Discoveries in the Nazca Lines
Using AI, researchers identified 1,309 potential geoglyphs, later confirmed through ground surveys approved by the Peruvian Ministry of Culture. This led to the discovery of 303 new relief-style geoglyphs in just six months—doubling the known figures in the region. The pace of discovery was 16 times faster than traditional methods, which had averaged just 1.5 finds per year.
The team used convolutional neural networks (CNNs), a deep learning model suited for image analysis. To address the limited training data, they leveraged pre-trained models on natural images. Instead of standard bounding-box detection, they used gridded classification to produce continuous probability maps at 5-meter resolution, improving geoglyph localization.
These discoveries offer new insights into the ritual life of the Nazca and Paracas cultures. Relief-type geoglyphs typically depict humans, llamas, and severed heads—likely reflecting daily life and small-group rituals. Their placement near ancient paths suggests they were meant for close-up viewing. In contrast, larger line-style geoglyphs, often showing wild animals, appear near extensive line networks, indicating community-wide pilgrimage or ceremonial use. This contrast helps archaeologists better understand Nazca society and its spiritual practices.
Small Things Considered
While AI is great at identifying and mapping archaeological structures, it is also used to reconstruct destroyed or incomplete sites. AI-powered platforms created by companies such as HeritageBuilder Tech leverage 3D scanning, photogrammetry, and historical information to generate rich digital twins of archaeological sites. Such virtual reconstructions enable researchers as well as the public to experience ancient habitats using virtual and augmented reality, and to save cultural heritage and share it globally.
In addition to the Nazca Lines, AI is also being used at archaeological sites globally to reveal and protect cultural heritage, as evidenced by the following examples:
1. Central American Mayan Ruins: AI-enhanced LiDAR (Light Detection and Ranging) scanning has been employed to create detailed maps of dense jungle areas in Guatemala, uncovering lost Mayan cities, roads, and buildings. With its ability to cut through dense foliage, LiDAR in combination with AI analysis identified tens of thousands of unknown buildings, revolutionizing our appreciation of Mayan urban design and population numbers.
2. Saruq Al Hadid, UAE: In 2023, scientists employed AI to study satellite imagery of this 1270–800 BCE copper-smelting site to recognize patterns of artifacts and pinpoint high-priority excavation locations. This method reduced disruptive excavation while optimizing efficiency of finds.
3. Pompeii, Italy: 3D modeling and AI-powered photogrammetry have been employed to rebuild damaged buildings and artifacts, producing digital twins enabling virtual exploration of the ancient city. They assist in research as well as public outreach, preserving the site for future generations to benefit from it.
4. Mesopotamian Floodplains: The Endangered Archaeology in the Middle East and North Africa (EAMENA) initiative utilizes AI to identify archaeological features based on satellite imagery, with up to an 80% success rate in identifying features such as burial mounds. This aids in prioritizing locations to be protected against looting and urban development.
These instances demonstrate how AI's capacity to process large datasets like satellite data, drone imagery, and LiDAR scans allows archaeologists to investigate inaccessible or covered-up locations non-invasively, conserving time and resources while safeguarding delicate remains.AI algorithms process structural elements and architectural designs to forecast missing components of ruins, creating a more comprehensive image of lost civilizations. The technology is also used to monitor and counteract damage by environmental agents such as climate change, which threatens to increase damage to vulnerable spots like those about which you were writing.
ScriptDecoder Technologies: Athens' Digital Linguistic Revolution
AI's transformative power extends well beyond site mapping and reconstruction to ancient script decipherment—a famously difficult task owing to fragmentary texts, vanished languages, and lack of bilingual references.
Headquartered in Athens, Greece, ScriptDecoder Technologies is leading the development of using AI to streamline the analysis of ancient manuscripts and inscriptions. Their digital solutions use optical character recognition (OCR) and natural language processing (NLP) specific to ancient scripts to allow automatic transcription, pattern identification, and preliminary analysis of historical writings. Such technologies greatly minimize labour that was once necessary for analysing scripts and create new doors of opportunity for educational enhancement and academic collaboration.
DeepMind's Ithaca: Restoring and Contextualizing Ancient
Google DeepMind’s Ithaca system marks a groundbreaking advance in epigraphy—the study of ancient inscriptions. Trained on a vast dataset of over 78,000 Greek texts dating from 700 BCE to 500 CE, Ithaca excels in three major tasks: restoring damaged inscriptions with 62% accuracy, pinpointing their geographical origin among 84 ancient regions with 71% accuracy, and dating them to within 30 years of their actual creation.
Unlike traditional methods that rely heavily on expert interpretation and manual comparisons, Ithaca uses machine learning to uncover linguistic and contextual patterns across thousands of texts. Researchers have already begun using Ithaca to revisit and reinterpret key periods of Greek history. Its adaptable framework can also be trained on other ancient writing systems, such as Mayan glyphs or cuneiform.By open-sourcing Ithaca’s models and tools, DeepMind encourages global academic collaboration. Now, researchers around the world can harness the power of AI to unlock insights hidden for centuries in ancient scripts.
Deciphering the Indus Valley Script: AI’s Role in Unlocking an Ancient Enigma
For more than a century, the mysterious script of the Indus Valley Civilization has captivated archaeologists and linguists alike. Scattered across ancient seals, pottery, and tablets, these enigmatic symbols—etched by one of the world’s earliest urban societies—have resisted all attempts at translation. The challenge looms large: most inscriptions contain only five symbols, and researchers have yet to discover a bilingual “Rosetta Stone” to serve as a key.
In recent years, however, a new protagonist has entered the scene: artificial intelligence. Research centers in India and abroad have begun harnessing AI to breathe fresh life into the quest for understanding. Where human eyes might miss subtle patterns or feel overwhelmed by the sheer volume of data, machine learning algorithms excel. These digital detectives analyze thousands of inscriptions, identify recurring motifs, and reveal hidden structures within the script. For instance, symbols like the “fish” and “jar” frequently appear together, suggesting grammatical rules or semantic relationships that had previously gone unnoticed.
But the story doesn’t end with pattern recognition. Millennia of weathering have eroded many Indus inscriptions, creating gaps and ambiguities in the record. Researchers now train machine learning models on both complete and partial symbols, enabling them to predict and reconstruct missing portions of damaged texts. These reconstructions have already uncovered new symbol sequences and combinations, offering valuable clues to the underlying logic of the script. At the Tata Institute of Fundamental Research and the University of Washington, researchers have gone a step further. They use AI to analyze the spatial arrangement and distribution of symbols, uncovering evidence that the script follows structured rules similar to those found in natural languages. This supports the tantalizing theory that the Indus script may encode a proto-Dravidian language spoken by the civilization’s people over four thousand years ago. This technological approach has ignited excitement beyond academia. In 2025, the Tamil Nadu government announced a $1 million prize for anyone who could crack the code of the Indus script. The news electrified the global research community, drawing in not only linguists and archaeologists, but also AI experts and amateur codebreakers from around the world. Contributors launched open-source projects on platforms like GitHub, training convolutional neural networks on thousands of seal images, clustering symbols, and sharing their findings in real time.
Yet, for all these advances, the Indus script remains an enigma. The absence of a bilingual reference text and the brevity of the inscriptions still pose formidable obstacles. AI has illuminated patterns and revealed a structural logic, but a full translation remains just out of reach. The most promising path forward likely lies in a hybrid approach, where the computational power of AI joins forces with the interpretive skills of human experts—each bringing their unique strengths to the challenge. As the story unfolds, the hope remains that one day, through the combined efforts of technology and scholarship, the silent voices of the Indus Valley will speak once more.
Comparison of AI-Assisted Methods to Traditional Methods
The AI-driven archaeological site mapping tools and ancient script deciphering tools provide a number of benefits compared to conventional approaches:
- Speed and Scale: AI processes vast datasets rapidly, enabling discoveries and analyses that would take decades manually.
- Accuracy and Sensitivity: Machine learning can identify subtle features and partial texts that other observers may miss.
- Cost-Effective and Non-Invasive: AI minimizes extensive fieldwork and invasive digs, which not only save resources but also protect sensitive locations and records.
- Integration of Multimodal Data: AI is capable of integrating imagery, text metadata, and historical context to deliver an overall perspective
But AI augments, not displaces, human expertise. Archeologists, linguists, and historians continue to be necessary to interpret AI-derived data, confirm results, and place outcomes within larger cultural contexts.
The Future of Archaeology and Historical Linguistics using AI
The use of AI in archaeology and decipherment of ancient script marks an era of new exploration and insight. AI, in Nazca, has hastened identification of long-concealed geoglyphs, adding to what is known about ancient civilizations in South America. Simultaneously, technologies such as ScriptDecoder Technologies’ platforms and DeepMind’s Ithaca open up textual voices of extinct civilizations, allowing scholars to read and date what were once considered indecipherable writing systems.
With further development of AI technologies, their fields of application will broaden into multiple archaeological contexts and linguistic traditions, helping to ensure that humanity's cultural heritage is not just conserved, but richly understood as well. Such collaboration of human learning and artificial intelligence is bound to enlighten the past to an unprecedented degree of clarity and accuracy.
Reference:
1. Youtube Video: 300+ NEW Nazca Lines Discovered with A.I. Technology
2. Youtube Video: AI Uncovers Lost Secrets of the Ancient World
3. https://swisscognitive.ch/2024/10/31/5-archaeological-discoveries-made-by-ai/
4. https://www.ultralytics.com/blog/ai-in-archaeology-paves-the-way-for-new-discoveries
5. https://digitaldefynd.com/IQ/ai-in-archaeology-case-studies/
6. https://www.pnas.org/doi/10.1073/pnas.2407652121
