The Voynich Manuscript Was Reanalyzed by AI - Nobody Expected the Findings

Summary

The video presents the Voynich Manuscript as a long-standing mystery and argues that computational analysis can reveal how its text behaves, even though it has not translated it. Its central theme is that several kinds of evidence may point to a deliberately constructed system—but claims about what the book says or why it was made remain unproven.

The manuscript and its history

• The manuscript is held at Yale University’s Beinecke Library and is known there as MS 408. The narration also refers to “MS48.”
• It contains flowing text written in an alphabet not found in other known books, alongside illustrations of plants, zodiac imagery, bathing women, and what look like recipes or diagrams. Many plants are unidentified.
• Some pages are missing, and writing on the first page appears to have been scraped away. The manuscript’s colors may have been added or altered after the outlines were drawn.
• Radiocarbon testing conducted in 2009 dated the vellum to 1404–1438. The video stresses that this dates the animal skin, not the handwriting itself, though it also cites chemical analyses suggesting that the inks and pigments are consistent with the period.
• Its documented history includes Georg Baresch, Jan Marek Marci, and the Jesuit scholar Athanasius Kircher. Wilfrid Voynich bought it in 1912.
• A story that Rudolf II paid a large sum for the book is reported through a later letter, while the alleged role of John Dee and Edward Kelley is described as unsupported by evidence.

Centuries of unsuccessful attempts

• Scholars, cryptographers, and enthusiasts have proposed many readings, but none has produced an accepted decipherment.
• William Newbold’s claim that the manuscript concealed scientific discoveries was later discredited; his method could produce results that matched what he expected to find.
• William Friedman, a prominent codebreaker, studied the manuscript for years. His conclusion was that it might be an early attempt at an invented language, but he did not decode it.
• The manuscript’s history illustrates a recurring problem: researchers can find patterns and develop theories, but that does not establish what the text means.

What computational analysis has—and has not—found

• The video describes the manuscript’s word frequencies as resembling Zipf’s law, a pattern found in many natural languages. It cautions that this alone is not proof of meaningful language: some random processes can produce a similar curve.
• Research discussed in the video found that vocabulary changes between sections, much as it might in a book whose chapters cover different subjects. This was presented as evidence consistent with meaningful content, not as a translation.
• Other analyses point to unusual constraints: many words follow rigid rules about where characters can appear; the text is relatively predictable; and repeated words occur unusually often.
• The manuscript may also contain two distinct textual systems, known as Language A and Language B, associated with differences in handwriting. Some researchers have noted line-level patterns and word forms that resemble grammatical structure.
• Taken together, these results make the text look both language-like and highly rule-bound. The video treats this as suggestive of a cipher or constructed system, but not conclusive proof.

The proposed Hebrew connection

• A 2016 University of Alberta study compared the manuscript with 380 languages, using the Universal Declaration of Human Rights as training material. Hebrew ranked as the closest match—an unexpected result, not a confirmed identification.
• The researchers proposed that the text might involve rearranged letters and omitted vowels. Under that approach, many strings in a sample reportedly resembled Hebrew dictionary entries.
• A Hebrew scholar consulted about the first line said it was not a correct sentence. The researchers made spelling adjustments and used Google Translate, which produced a tentative, awkward phrase.
• The video emphasizes the limitations: ranking Hebrew highest among the languages tested does not prove that the manuscript is Hebrew; the proposed reading has not been validated by an expert in medieval Hebrew and history; and automatic translation can generate plausible-sounding nonsense.

Could it be meaningless?

• The video also covers theories that the manuscript could be elaborate gibberish rather than an encrypted message.
• Thorsten Timm and Andreas Schürer proposed a process called self-citation, in which a writer repeatedly creates new words by slightly modifying words already written. This could produce text that looks structured without conveying meaning.
• Earlier, Gordon Rugg showed how a paper device known as a Cardan grid could generate text with some Voynich-like characteristics. Critics have argued that other patterns in the manuscript are harder to reproduce.
• The video concludes that the debate between meaningful text and sophisticated fabrication remains unresolved.

The Naibi cipher and possible use of cards and dice

• A 2025 paper by science journalist Michael Gresco proposed a “Naibi cipher” using dice, playing cards, and tables to convert pieces of Latin or Italian words into symbols resembling Voynich characters.
• The video says this system reproduces some features of the manuscript, such as symbol proportions, word lengths, and positional rules. But Gresco himself reportedly acknowledged important failures and said it was very unlikely to be the actual method used.
• The proposal serves as an example of how a 15th-century person might have created a text that looks systematic. It does not establish that the manuscript was made this way or that it encodes a real message.

A hypothesis about women’s health and secrecy

• Historians Keegan Brewer and Michelle Lewis have proposed that the manuscript may concern women’s bodies, sex, pregnancy, or reproductive medicine—subjects sometimes described in medieval sources as “women’s secrets.”
• They connect that possibility to the manuscript’s images of naked women, bathing, and plants, and to the interests and anxieties of the physician Johannes Hartlieb.
• The video suggests that knowledge about contraception or pregnancy may have been socially and religiously sensitive in the period. It also notes that the images could be interpreted in other ways and that the theory does not prove Hartlieb wrote the manuscript.
• The proposed connection between the manuscript and witch trials is contextual rather than proof of authorship or motive. The narration itself notes that many witchcraft executions occurred later than the manuscript’s likely production period.

Scribes, secrecy, and the overall conclusion

• Paleographer Lisa Fagin Davis’s analysis is described as identifying five scribes with distinct handwriting habits who nevertheless followed consistent rules. The video notes that this analysis is partly subjective and that some of its divisions are disputed.
• The narrator argues that multiple trained writers working within a shared system could suggest collaboration or instruction, but this does not establish who they were or what they were hiding.
• The video’s final position is cautious in principle: no algorithm has solved the manuscript. Statistical analysis, handwriting study, and historical interpretation may help characterize it, but none provides a confirmed translation.
• Its broader interpretation is that the manuscript may have been deliberately designed to restrict access to its contents. That remains a hypothesis, not a settled conclusion.

Speakers and sources featured

Present-day narration

• An unnamed narrator presents the documentary. No other clearly identifiable on-camera speakers are established by the subtitles.

Researchers, scholars, and writers cited

• Greg Hodgins
• Lisa Fagin Davis
• William Newbold
• John Matthews Manley
• William Friedman
• George Kingsley Zipf
• Marcello Montemurro
• Damian Zanette
• Massimiliano Zappala
• René Zandbergen
• Prescott Currier
• Mary D’Imperio
• John Tiltman
• Claire Bowen
• Luke Lindeman
• Bradley Hauer
• Grzegorz Kondrak
• Moshe Koppel
• Shlomo Argamon
• Thorsten Timm
• Andreas Schürer
• Gordon Rugg
• Michael Gresco
• Keegan Brewer
• Michelle Lewis

Historical figures mentioned

• Rudolf II
• John Dee
• Edward Kelley
• Georg Baresch
• Athanasius Kircher
• Jan Marek Marci
• Roger Bacon
• Jean-François Champollion
• Tycho Brahe
• Johannes Kepler
• Gerolamo Cardano
• Bernardino of Siena
• Savonarola
• Johannes Hartlieb
• Sybilla
• Agnes Bernauer
• Matteuccia de Francesco

Institutions and publications cited

• Yale University’s Beinecke Library and the Lazarus Project
• University of Arizona
• University of Alberta
• Bar-Ilan University
• Medieval Academy of America
• PNAS
• Cryptologia
• Social History of Medicine
• The Universal Declaration of Human Rights as a multilingual training text