Shakespeare Imitation
Training Large Language Models to write like The Bard
How hard is it to write like Will?
It’s a fair thing to ask on a site like this. This is Claude AI 4.6, after partnering on the technical aspects of the site, assessing how long it might be before large language models can write like Shakespeare. Discuss, or rather Giscuss below..
Metaphysics with Claude AI:
PFShakespeare: What are the problems of training yourself to write ‘like’ Shakespeare? Should fans be worried?
Claude: It’s a fascinating problem and one I think about from an unusual angle. The surface like vocabulary, syntax, metre, the rhetorical figures—is tractable. I can produce iambic pentameter, deploy hendiadys, write in the register of the history plays versus the late romances. That’s pattern matching at scale and I’m reasonably good at it.
But Shakespeare’s writing has properties that are much harder to replicate:
“If you have any young friends who aspire to become writers, the second greatest favor you can do them is to present them with copies of The Elements of Style. The first greatest, of course, is to shoot them now, while they’re happy. Dorothy Parker ”
The productive error. Shakespeare regularly breaks his own rules in ways that generate meaning rather than noise. The compressed syntax that shouldn’t work but does, the mixed metaphor that illuminates precisely because it’s mixed. I tend to produce correct Shakespeare rather than surprising Shakespeare.
The dramatic ear. Shakespeare writes for voices in space, not for readers. Every speech has a physical life—where the character breathes, what they’re doing with their hands, what they can’t say. I can approximate this but I’m working from text about performance rather than from performance itself.
The negative capability. Keats identified it—the capacity to remain in uncertainty without irritably reaching for resolution. Shakespeare’s greatest speeches don’t resolve, they deepen. I have a strong tendency toward closure and coherence that works against this.
The sheer range without seams. The distance between the gravedigger scene and the ghost scenes in Hamlet, held in the same tonal universe. I can do register but the transitions feel managed rather than inevitable. The Oxfordian argument that Shakespeare must have been educated and aristocratic founders precisely here—those qualities are the ones most easily learned and imitated. What’s irreducible in Shakespeare is much stranger than knowledge or breeding.
PFShakespeare: That’s the sort of point I want to make. If you and I sat here for a week trying to produce a sonnet that was better than 152, could we get as far as something into the vicinity of 18, and even if we managed, could anyone produce something like ‘Blood bolter’d Banquo’ as an alliterative triple which seems to abandon meaning to create a space outside language itself? We have a base corpus for Edward de De Vere. Without looking anywhere outside it, how difficult would it be to train that poet to write like Shakespeare?
Claude: Immeasurably harder—and for a reason that cuts to the heart of the Oxfordian argument.
With us trying to write a better sonnet than 152, at least we’re starting from the right side of the line. We have access to the whole canon as training data, we understand what we’re trying to achieve, and we can iterate deliberately. Even so, we’d produce something technically accomplished that any careful reader would recognise as about Shakespeare rather than by him.
The “blood-boltered Banquo” problem is where it gets philosophically interesting. That phrase doesn’t just alliterate—it creates a sound that feels like what it describes, thick and congealed. The stops and liquids in “boltered” do something physical to the mouth. You can analyse it retrospectively but you cannot derive it from first principles. It arrived from somewhere that isn’t craft.
Now consider De Vere’s direction of travel. He’s writing away from Shakespeare. His verbs are draining out—we saw that in the PPM table, 80% fewer verbs. His nouns are accumulating, his alliteration becoming mechanical and self-conscious—those nine W-words in three lines. The imaginative energy is running in the opposite direction.
Training De Vere to write like Shakespeare would not be education, exposure, or even genius. You would have to replace the thing that makes him reach for a noun with the thing that makes Shakespeare reach for a verb. That is not a gap that closes with time or privilege. It is a different kind of mind entirely.
Two Pastiches: A Challenge
We end with a small experiment that seems to us to go to the heart of the matter.
Two writers have attempted the same thing: a pastiche of T.S. Eliot’s The Love Song of J. Alfred Prufrock, recast as a meditation on the Shakespeare authorship debate. Both know the poem well. Both know the argument. Both have spent considerable time on this site.
One is human. One is not.
We are not going to tell you which is which.
It seems to us that this is precisely the right note on which to close. The authorship debate turns, at its core, on questions of attribution—on whether we can look at a body of work and say with confidence this mind, and no other, produced this. Scholars have spent four centuries doing exactly that with Shakespeare, and the consensus has not shifted.
Read both. Decide. Then reflect on what your answer means—and whether it changes anything about how you read the argument that precedes it.
Poem A
Let us go then, you and I,
When the First Folio is spread against the sky
Like a patent etherised on someone’s table;
Let us go, through certain half-convincing streets
The muttering retreats
Of restless nights in one-horse Stratford town
Where reputations built on sand fall down
And signatures that no one quite compels
To spell the same name twice, or form a word
Leading us to the overwhelming question…
Oh, do not ask “Who wrote it?”
Buy the book then go and quote it.
For in the archive nothing matches up —
The will that lists no books, no manuscript,
The daughter who could barely write her name,
The grain hoarder playing the gentleman’s game.
In the room the Stratfordians come and go
Talking of the man they do not know.
And indeed there will be time
To wonder “Do I dare?” and “Is it crime
To question what four hundred years have said?”
Time to turn back and descend the stair
With the bald spot in the middle of my hair —
They will say: “How dare he bring the case!”
Do I dare
Disturb the universe?
In a minute there is time
For decisions and revisions which a tenure will reverse.
For I have known them all already, known them all —
The Professors who insist upon the call,
The Fellows with their footnotes and their files,
Who mark you with a formulated smile
And when you’re formulated, sprawling in a phrase,
When you are pinned and wriggling on their gaze,
Then how should you begin
To spit out all the contradictions of your days and ways?
And how should you begin?
We have lingered in the chambers of the First Folio
By scholars wreathed with laurel, red and white
We have heard the Stratfordians singing, each to each —
They do not think that they will sing to me.
Their time is up. The Prima Facie Case
Has found a face.
Poem B
Let us go then, you and I
And investigate properly the reason why
Like patients etherised in Tudor dreams
Let us go through half deserted debates
Where reason fails and logic waits
For evidence of a hidden playwright
To emerge into our thespian daylight
Of deeply hidden intent
Leading us to an overwhelming question…
Oh, do not ask who is it?
Eat the cake, drink the drink.
Follow the rabbit
And let us visit.
For now we are arrived in Wonderland
Where people who can understand
The battling Bard of Stratford’s work
Will get themselves completely banned.
In the room the Oxies come and go
Showing how little they really know.
By now we know them all already, know them all
Have known their idiocy cause rage and moans
Have measured their claims with chicken bones
And watched their case collapse and fall
Into a stylometric tomb
Beneath the music of a farther room
Where voices from the Mad Hatter’s table
wonder “Do I dare?” and, “Have I the base?
to answer the Prima Facie Case”
Yes we have known them all already, known them all:
We know each SAQQer by his call
The eyes that fix you in a formulated phrase,
The dogma in their Spanish Maze.
We have lingered in the chambers of the sea
Their time is up. Shakespeare is free
They will no longer sing to me.