The worry about AI writing is that a machine can take in human work and hand back something built from it. That fear is old. Long before machines, people made new texts out of the texts they inherited. This walks through three of the oldest examples, and separates two ideas we tend to blur. It ends where universities now stand: judging not whether a text is original, but whether its maker was honest about how it was made.
Module 01
Two words we blur
In this module: what separates plagiarism from intertextuality, and why the ancient world had both
Before the AI question makes sense, two ideas need separating. One names a fault. The other names how almost all writing has always worked.
One is deception; the other is method
Plagiarism
Passing off another’s words or ideas as your own, without credit. The fault is not the reuse. It is the deception. Plagiarism needs an intent: to make borrowed work look like something you produced yourself.
A moral judgement. It hides its source on purpose.
Intertextuality
Reworking earlier texts to make new meaning. The literary theorist Julia Kristeva coined the term in 1966 for the way every text is woven from others. The reuse is open, and the aim is to transform, not to conceal.
A description of how texts get made. It builds on its sources in plain sight.
The word intertextuality is Kristeva’s, from her 1966 essay “Word, Dialogue and Novel”. Hold the difference in mind: one term is about honesty, the other is about method. They are not opposites, and a single act can raise both questions at once.
Antiquity had the concept, and a word for it
It is a myth that the ancient world copied freely because it had no rules. Ancient writers named and shamed literary theft. The Roman architect Vitruvius, writing around 30 to 15 BCE, records one such case. The scholar Aristophanes of Byzantium exposed poets at a contest in Alexandria who had recited other people’s work as their own.
Vitruvius, De architectura, Book VII, preface. So the ancient world could tell theft from tradition. What differed was the default: reuse was normal, and authorship was rarely treated as private property.
Where the word comes from
Guess before you look. The English word plagiarism comes from a Latin word. What did that Latin word first mean?
A kidnapper. The Latin plagiarius meant someone who abducts another person’s slave or child. The Roman poet Martial, writing around 86 to 103 CE, borrowed the word as a metaphor: a rival who passed off Martial’s poems as his own had, in effect, kidnapped them. The literary sense stuck, and it still carries that charge of stealing a person, not just a phrase.Oxford Classical Dictionary, “Plagiarism in Latin literature”; Martial, Epigrams 1.52. Etymology via etymonline.
Why the line shifted
Ancient authors mostly worked in a world where inspiration was seen as an outside gift, even a divine one, rather than personal property. So what we would call plagiarism today was often something else: a sophisticated tradition of reuse. Building on inherited stories was the expected craft, not a failure of it.
To see how far that went, look at the texts at the very root of the tradition. Each one was made by reworking an older one.
Module 02
Built on older texts
In this module: three foundational texts, each made by reworking an earlier one
The texts at the root of Western culture are not original in the modern sense. Each was assembled from stories that came before. Here are three, oldest first.
The flood: Genesis and Gilgamesh
Older
The Epic of Gilgamesh
A man is warned of a coming flood. He builds a great boat, loads it with living things, rides out the deluge, then sends out birds to find dry land. Many gods drive the story.
The Babylonian standard version was compiled by the scribe Sin-leqi-unninni from older Sumerian poems, roughly 1300 to 1000 BCE.
Later
Genesis, the story of Noah
Nearly the same plot: a warning, an ark, the animals, the flood, the birds sent out. The authors did not copy blindly. They rewrote the polytheistic tale into one about a single God, a covenant and a moral order.
Same structure, a different worldview. The reuse carries a new meaning.
The flood parallel is standard scholarship on the ancient Near East. The point is not that Genesis stole; it is that a shared story was reworked to say something the older version did not.
The same epic, a different god on top
The Enuma Elish is the Mesopotamian creation epic, seven tablets long. The surviving copies date to about 1200 BCE, and are themselves copies of an older work. Click each city to see who plays the hero.
Click a city. The plot barely changes; the name at the top does.
Two Gospels copying a third
Matthew, Mark and Luke tell overlapping stories, often word for word. Most scholars explain this with one relationship. Click each to see it.
Click a Gospel to see how it relates to the others.
Reuse was the method, not a lapse
Across all three cases, the writers built on inherited material and bent it to a new purpose. Every past text was treated as a starting point for a later one. The reuse of existing stories was not a lack of imagination; it was a deliberate way of building on the shared literature of a culture.
Hold that in mind for the modern question. A language model does something that looks a lot like this, at a scale no scribe could reach.
Module 03
The machine, and the human it copies
In this module: what a language model does, and whether the worry is new
Now the modern question. A large language model reads across enormous amounts of human text and predicts likely next words. Set beside this history, does it do something new, or something very human at industrial scale?
What a language model actually does
A model has no store of passages to lift. It learns relationships between words across its training text, then makes new text by predicting what tends to come next. In a mechanical sense that is intertextuality without an author: recombination of inherited material, faster than any scribe.
Plagiarism, intertextuality, or neither?
Three cases. For each, decide what it is. There is a defensible answer, and the interesting part is why.
Case one. A student pastes an AI-written essay under their own name and submits it as their own work.
Plagiarism. The deception is the point: the work is passed off as the student’s own. The tool barely matters here; a ghost-written essay bought from a person would be the same fault.
Case two. The authors of Genesis retell the Gilgamesh flood, openly within their own tradition, to teach a different theology.
Intertextuality. Open reuse that transforms meaning. No one is deceived, and the whole point is the new frame placed on an old story.
Case three. A language model writes a paragraph by predicting words from patterns across millions of texts, with no single source and no intent.
Neither, cleanly. There is no author to deceive and no single source being reworked. The old categories were built for humans. The machine sits outside them, which is exactly why it unsettles the rules.
Two things the history makes clear
The tide is old
Learning from existing work and recombining it is not a modern crime. It is how culture has always moved. Rules can shape how the tools are sold and used. They are unlikely to stop machines learning from what humans have already written.
The human source goes missing
Plagiarism is an intentional deception. A model does not intend anything; it predicts. Judging a machine by a rule written for a lying human misfits the case. The fault, where there is one, sits with the person who hides how the work was made.
Why this matters
The history does not tell universities to drop academic integrity. It relocates the question. For a few centuries we judged the originality of the text. That test breaks when a machine can produce fluent, source-less prose on demand. What survives is the question antiquity actually cared about: not is this text new, but was its maker honest about how it was made? The shift is from policing originality to expecting transparency. The essay behind this deck ends on the open one to take to a room, and to students themselves: how should universities redefine misconduct to account for machine-assisted intertextuality?
Sources used. Plagiarism and intertextuality: the term intertextuality was coined by Julia Kristeva, “Word, Dialogue and Novel”, 1966. Ancient attitudes: Vitruvius, De architectura, Book VII, preface, c. 30 to 15 BCE (the Aristophanes of Byzantium anecdote); “Plagiarism in Latin literature”, Oxford Classical Dictionary; the word plagiarius and Martial, Epigrams 1.52, c. 86 to 103 CE, with etymology confirmed via etymonline. Flood narratives: the Babylonian standard version of the Epic of Gilgamesh compiled by Sin-leqi-unninni from older Sumerian poems, c. 1300 to 1000 BCE; the Genesis and Gilgamesh flood parallel is standard scholarship on the ancient Near East. Enuma Elish: seven tablets, surviving copies c. 1200 BCE and themselves copies of an older work, with the copy found at the city of Ashur placing the Assyrian god Assur in the hero role held by Marduk in the Babylonian form (World History Encyclopedia and Encyclopaedia Britannica, “Enuma Elish”, accessed August 2026). Synoptic Gospels: Marcan priority, the majority scholarly view that Matthew and Luke used the earlier Gospel of Mark, drawing also on shared oral tradition (Encyclopaedia Britannica; Bart D. Ehrman). Language-model mechanism: the standard description of next-token prediction. The framing and the closing question are adapted from an essay in the CDU workshop materials on LLM anxiety and intertextuality. One claim from that essay, that the Enuma Elish borrowed specifically from the earlier Anzu myth, was left out: the sources consulted confirm the epic reworks an older original and that the Assyrian copy swapped in Assur, but did not confirm the Anzu detail.