Introduction
The integration of artificial intelligence into archaeological analysis has yielded a breakthrough in our understanding of some of the most historically significant manuscripts ever discovered: the Dead Sea Scrolls. According to a June 2025 report published in the peer-reviewed journal PLOS One, researchers have used a machine learning model, paired with modern radiocarbon dating, to reassess the age of various fragments of the Dead Sea Scrolls. This dual-methodology approach has resulted in revised chronological estimates, placing some scrolls up to a century older than previously believed. The implications of this reevaluation extend beyond textual dating. They influence theological studies, cultural history, and methods of digital manuscript authentication.
Historical Context of the Dead Sea Scrolls
The Dead Sea Scrolls were discovered in 1947 by Bedouin shepherds in the Judaean Desert near Khirbat Qumran. This accidental find led to one of the most substantial archaeological endeavors of the twentieth century. The fragments, numbering in the thousands and representing approximately 1,000 distinct manuscripts, were recovered from 11 caves situated near the Dead Sea in what is now the West Bank. Most of the texts are written in Hebrew, with some in Aramaic and a smaller portion in Greek. The scrolls include biblical texts, apocryphal writings, and sectarian documents that offer insight into the beliefs and practices of Second Temple Judaism.
The scrolls have long been studied using paleography, the analysis of ancient handwriting. Based on this traditional approach, scholars placed the scrolls within a broad window spanning the third century BCE to the second century CE. This timeframe, however, has been subject to dispute due to limited data and varying interpretive methodologies.
Modern Radiocarbon Dating and Its Limitations
Radiocarbon dating, developed by chemist Willard Libby in the late 1940s, has been instrumental in authenticating and dating organic archaeological materials. This method measures the ratio of radioactive carbon isotopes within a given sample to estimate age. While highly effective, it comes with significant drawbacks. The process is destructive. It requires the physical removal of material from the specimen, which is then chemically analyzed and destroyed in the process. Given the fragile and historically irreplaceable nature of the Dead Sea Scrolls, this presents ethical and material challenges.
Compounding the problem is contamination. Specifically, castor oil was used in the mid-20th century to enhance legibility of the text, which introduced modern organic material that skewed the radiocarbon dating results toward more recent timeframes. Consequently, the 1990s-era estimates were called into question for their possible misrepresentation of the actual creation period of the texts.
Introduction of Enoch: AI-Powered Analysis
To address these limitations, a team of researchers led by Mladen Popović at the University of Groningen introduced an artificial intelligence model named Enoch. This machine learning model was trained on high-resolution imagery of Dead Sea Scroll fragments that had undergone modern radiocarbon testing. Enoch was developed not only to replicate human paleographic analysis but to exceed it in precision and scalability. By isolating subtle features in handwriting, such as stroke width, curvature, ink deposition, and letter formation patterns, the AI model aimed to determine the time period of script origin with high fidelity.
In its validation phase, Enoch was presented with imagery of scrolls already dated by updated carbon 14 methods. The model was not told the actual dates but was tasked with independently producing age estimates. Remarkably, it matched the radiocarbon results with 85 percent accuracy. In many instances, it provided even narrower age windows than carbon dating could offer. Following validation, Enoch was applied to 135 scrolls not previously subjected to carbon analysis. Human researchers then evaluated the AI-generated age estimates. The model produced realistic chronological placements in 79 percent of cases, marking a significant achievement in automated manuscript analysis.
Key Findings and Revisions
The study’s most prominent conclusion is that several scrolls are between 50 and 100 years older than previously estimated. One such fragment containing text from the Book of Daniel, formerly believed to originate in the second century BCE, is now dated to the time of its likely authorship in the fourth century BCE. This alignment has profound implications for both textual authenticity and historical contextualization. Another manuscript containing verses from Ecclesiastes was revised from an estimated origin in the mid-second century BCE to between 300 and 240 BCE.
These revisions not only affect our understanding of when specific texts were written but also reframe discussions surrounding the dissemination, copying practices, and audience of the documents. If texts existed earlier than assumed, their influence and usage across communities must be reevaluated accordingly.
Theoretical and Practical Implications
The success of the Enoch model raises possibilities for similar applications in other historical corpora. Manuscripts written in Syriac, Latin, Greek, and Arabic, which lack reliable chronological documentation, could benefit from this type of non-destructive analysis. Further, the integration of AI in manuscript dating could eventually supplant destructive chemical methods, preserving the integrity of rare documents.
The implications also extend to theology and literary history. As one of the oldest surviving witnesses to the Hebrew Bible, the Dead Sea Scrolls inform not only textual variants but also religious doctrine, canonical development, and intertextual influences. Revising their dating introduces potential recalibrations in the timeline of biblical transmission.
Limitations and Scholarly Reception
Despite its promise, the Enoch model is not without limitations. Its success relies on the quality and quantity of training data, which, due to the finite number of radiocarbon-dated scrolls, remains constrained. Additionally, manuscript conditions such as fading ink, physical degradation, and inconsistent handwriting styles introduce variables that could compromise machine inference.
External scholars have weighed in with cautious optimism. Charlotte Hempel from the University of Birmingham highlighted the value of combining AI and radiocarbon data for cross-calibration. She emphasized the model’s apparent ability to provide narrower date ranges than traditional methods. Lawrence Schiffman of New York University echoed the sentiment, noting that while the approach is promising, its efficacy on undated texts remains to be fully verified. Brent Seales from the University of Kentucky commended the rigor of the study but cautioned against viewing AI as a complete replacement for radiocarbon dating at this stage. He positioned the model as a complementary tool that enhances the broader analytical ecosystem.
The Road Ahead
Popović and his team have stated their intent to continue refining the Enoch model. Enhancements could include broader training datasets, improved neural architectures, and contextual reinforcement using metadata such as material composition or geographic provenance. Should such refinements prove successful, Enoch could be deployed in a variety of historical and forensic applications. Its non-destructive nature makes it especially suitable for cases where sample preservation is paramount.
Moreover, the researchers envision AI as part of a new scholarly paradigm where human expertise and machine learning work in tandem. Instead of supplanting human judgment, AI models like Enoch serve as high-speed, high-consistency tools that can pre-process or validate human hypotheses.
Conclusion
The fusion of machine learning with archaeological methodology has enabled a reevaluation of the Dead Sea Scrolls’ chronology. Through the Enoch AI model, researchers have not only validated but also refined prior dating efforts, uncovering new dimensions of these ancient texts. This represents a paradigmatic shift in the authentication of historical documents and reinforces the value of interdisciplinary approaches that blend digital innovation with traditional scholarship.
With continued development, AI-based paleography could serve as a gold standard for non-invasive manuscript dating, ensuring both accuracy and preservation. As we stand on the frontier of digital archaeology, Enoch exemplifies the capacity of artificial intelligence to deepen our understanding of the ancient world.
Works Cited
Popović, Mladen, et al. “Artificial Intelligence Analysis of Dead Sea Scroll Manuscripts.” PLOS One, 5 June 2025.
Prisco, Jacopo. “AI Analysis of Ancient Handwriting Gives New Age Estimates for Dead Sea Scrolls.” CNN, 7 June 2025, www.cnn.com/2025/06/07/dead-sea-scrolls-ai-study.
Seales, Brent. Personal correspondence. University of Kentucky.
Hempel, Charlotte. Personal correspondence. University of Birmingham.
Schiffman, Lawrence. Personal correspondence. New York University.