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

Summary

The video presents the Voynich Manuscript as a 15th-century book written in an unknown script and illustrated with plants, zodiac imagery, bathing women, and other puzzling scenes. It argues that AI has not translated the manuscript, but has helped assess what kind of text it might be and what it may have been intended to protect.

The manuscript and its history

• The manuscript has about 240 pages, written in a consistent, unfamiliar alphabet. Its smooth handwriting and organized-looking paragraphs suggest deliberate composition rather than casual scribbling.
• The illustrations include unidentified plants, star and zodiac diagrams, women bathing in connected pools, and a series of large circular diagrams. Some pages are missing, and the video says a name was once scraped from the first page.
• Radiocarbon tests on the parchment reportedly dated it to 1404–1438, supporting the conclusion that it is genuinely medieval rather than a modern fake.
• The video traces the manuscript through several owners: Emperor Rudolf II, alchemist Georg Baresch, scholar Johannes Marcus Marci, and book dealer Wilfrid Voynich. None of the people described in the account managed to read it.
• Earlier efforts at decipherment produced confident but disputed claims. The video describes William Newbold’s interpretation as a misreading of ink cracks and notes that codebreaker William Friedman ultimately believed the text might represent an invented language.

What pattern analysis found

The video says researchers shifted from asking “What does it say?” to asking “What kind of system could have produced it?”

• The text reportedly follows Zipf’s law, a statistical pattern common in natural languages.
• Its vocabulary varies across sections, which may indicate that different parts concern different subjects.
• Other features seem unlike ordinary language: most words allegedly follow strict rules about where letters can appear, the next character is unusually predictable, and words sometimes repeat several times in a row.
• The manuscript also appears to contain two textual systems, associated with different handwriting styles, and its words behave differently at the beginnings and ends of lines.
• Together, these findings are presented as evidence that the text has both language-like and mechanically constrained features. The video suggests this could fit a cipher, but says the same patterns might also arise from a carefully constructed fake.

AI language comparison and the Hebrew proposal

• In the account of a 2016 study, researchers trained an AI system using the Universal Declaration of Human Rights translated into many languages. The system then compared the manuscript with those languages.
• Hebrew was reportedly its closest match. The researchers hypothesized that the text’s letters might have been rearranged and many vowels omitted.
• After rearranging letters in the opening line, the researchers reportedly found that many resulting forms resembled Hebrew dictionary words. A machine translation produced an awkward sentence beginning with “She.”
• The video treats this as a possible clue, not a decipherment. It notes that the result was not a proper Hebrew sentence and that a qualified specialist would need to assess the proposal.

A possible way the text could have been encoded

• The video describes a 2019 “self-citation” proposal: a scribe could generate a language-like but meaningless text by repeatedly copying nearby words and changing them slightly.
• It also describes a 2025 study proposing the “Naibbe cipher.” In this method:
  1. Start with a sentence in Latin or Italian.
  2. Roll a die to determine how to divide each word into short letter groups.
  3. Draw a playing card to select one of several substitution tables.
  4. Convert the letter groups into symbols resembling the manuscript’s script.
• The video says text generated this way resembles the manuscript in symbol frequencies, word lengths, and other constraints. It presents this as evidence that a real message could have been hidden with methods available in the 15th century—not proof that this exact cipher was used.
• The connection to dice and playing cards is framed as significant because the video associates those objects with moral and religious opposition in medieval Europe.

The proposed subject: women’s medical knowledge

The video gives greatest weight to a theory that the manuscript may contain knowledge about women’s bodies, reproduction, and medicine.

• Historians Keegan Brewer and Michelle Lewis are cited for focusing on the manuscript’s bathing women and imagery related to female anatomy.
• The video connects the illustrations to medieval “women’s secrets”—knowledge about sex, pregnancy, childbirth, and remedies that could be considered dangerous or forbidden.
• It proposes that the plants might represent medicinal remedies, the zodiac imagery might relate to pregnancy and birth, and the nine-part circular illustration might refer to reproductive anatomy. These are interpretations, not confirmed readings of the pictures.
• The video also links the theory to physician Johannes Hartlieb, who wrote about plants, astrology, baths, magic, and women, and reportedly advocated concealing certain reproductive recipes.
• The opening word “She,” from the proposed Hebrew interpretation, is presented as suggestive but remains uncertain because the line has not been reliably translated.

The video’s conclusion

• The manuscript’s five distinct scribal hands, as described by handwriting expert Lisa Fagin Davis, are presented as evidence that several trained writers shared a system rather than one person improvising nonsense.
• The video connects the proposed subject matter to the dangers faced by people who held or transmitted women’s medical knowledge, including midwives, in an era of witch hunts.
• It concludes that the manuscript was likely encoded deliberately, perhaps to protect its contents and its makers. However, it also acknowledges that this is not a completed translation. The video’s central claim is therefore a hypothesis: the book may be a real, intentionally concealed text, but its meaning remains unresolved.

Speakers and sources featured

The subtitles do not identify distinct interview speakers. They appear to contain documentary narration and references to the following people and sources:

• Narrator: Unidentified voice-over.
• Researchers and scholars: Marcelo Montemurro, Damian Zanette, Massimiliano Zattera, Bradley Hauer, Grzegorz Kondrak, Torsten Timm, Andreas Schinner, Mary D’Imperio, Michael Greshko, Keegan Brewer, Michelle Lewis, and Lisa Fagin Davis.
• Historical figures discussed: Wilfrid Voynich, Rudolf II, Georg Baresch, Athanasius Kircher, Johannes Marcus Marci, William Newbold, William Friedman, Johannes Hartlieb, Bernardino of Siena, Agnes Bernauer, and Sibylla.
• Institutions and publications mentioned: Yale’s Beinecke Rare Book and Manuscript Library, the University of Arizona, the University of Alberta, the U.S. National Security Agency, Cryptologia, and Social History of Medicine.