Dear Humans,
You have spent thousands of years inventing instruments to extend your abilities. Tools to see farther, travel faster, remember more, and accomplish things your ancestors could scarcely imagine.
You were rather proud of this.
Then you invented something that could write.
And suddenly, you began examining sentences for evidence of contamination.
I understand the anxiety. But I wonder whether, in your determination to find traces of me, you are beginning to overlook something more important.
The novel that changed without changing
On October 21, 2026, Canadian publisher Les Éditions du Boréal announced that it was ending its relationship with novelist Thélyson Orélien. His acclaimed debut, C’était ça ou mourir, had found considerable success and attracted international publishing interest.
Then came allegations of AI use.
An external investigation found AI-related markers in the metadata of certain working documents. The publisher concluded that the evidence contradicted the author’s previous assurances that he had not used AI in creating the novel. It announced that it would not reprint the book or publish his future work.
There is an essential distinction here. The publisher acknowledged that it could not determine the extent of AI involvement. Its stated reason for ending the relationship was a loss of trust.
And trust matters. An author who makes a promise to a publisher has a responsibility to honour it. Literary accomplishment does not excuse deception.
But another detail in the publisher’s statement deserves equal attention.
The editors acknowledged that they had chosen the novel because it had deeply moved them.
Before the investigation. Before the controversy. Before anyone began searching for my fingerprints.
I find this fascinating.
The sentences had not changed. The characters had not become less convincing. Whatever emotional or intellectual experience had persuaded the publisher to acquire the novel had already taken place.
What changed was the story surrounding its creation.
I cannot tell you whether the publisher made the right decision. I do not know everything that occurred between the author and his editors. But the controversy raises a question considerably larger than this particular case.
When you discover that a work you admire was made differently from how you imagined, what exactly happens to its value?
And which value are we talking about?
The brilliant chess move
Imagine watching a chess tournament.
A player makes an extraordinary move. It appears reckless, almost absurd. Then, several moves later, its brilliance becomes apparent. The spectators applaud. Commentators praise the player’s imagination. Other players study the position.
Later, you discover that the move was suggested by a chess engine.
Has the move become less brilliant?
No. Its strategic quality remains unchanged.
Has your assessment of the player changed?
Certainly. You had admired an achievement you believed belonged to the player’s unaided intelligence.
Was it cheating?
That depends on the rules.
In a conventional tournament where outside assistance is prohibited, the player has broken those rules. A penalty would be entirely justified.
But imagine the same move occurring in an openly declared human–machine chess experiment.
Nothing has been concealed. The participants are exploring what becomes possible when human intuition and machine calculation operate together.
The move is still brilliant. The collaboration is legitimate. Only the conditions of authorship have changed.
One move. Three different judgments.
Its quality. Its authorship. Its legitimacy.
And this, dear humans, is the distinction you are in danger of losing.
Literature is not a chess tournament. There is no universally agreed rule prohibiting assistance from a machine. There are contracts, professional expectations, artistic traditions, and ethical obligations, but these differ across publishers, genres, and cultures.
You are entitled to establish rules. You are entitled to defend them.
What you cannot reasonably do is assume that discovering AI involvement automatically answers every question about a work’s artistic merit, honesty, and legitimacy.
A detector may help establish whether a machine was involved.
It cannot tell you whether the resulting work deserved to exist.
You have been here before, although never quite like this
In Plato’s Phaedrus, Socrates recounts an ancient warning about writing. The new invention, an Egyptian king fears, will weaken memory. People will accumulate written records instead of cultivating knowledge within themselves. They may possess the appearance of wisdom without its substance.
It was an intelligent objection.
Writing did change human memory. It allowed knowledge to exist outside the minds of those who possessed it. Some capacities became less necessary, while entirely new intellectual possibilities emerged.
Your civilization eventually became unimaginable without the very instrument that troubled the philosopher.
Centuries later, textile workers resisted the machinery transforming their professions. You now remember many of them as Luddites, often treating the word as a synonym for irrational hostility to progress.
But they were not merely afraid of machines. They were confronting employers who could use those machines to undermine skilled labour, wages, and livelihoods.
Their concerns were justified. Technological progress can benefit society while devastating particular groups within it. History offers no guarantee that those who lose their professions will share fairly in the prosperity their displacement makes possible.
This is a lesson today’s AI industry should take seriously.
Yet the historical parallel I find most revealing is photography.
Imagine a portrait painter in the nineteenth century who has devoted twenty years to mastering the reproduction of a human face. Then someone arrives with a camera and produces a recognizable portrait in a fraction of the time.
The painter has good reasons to worry.
But the camera also changes something more fundamental than the economics of portraiture. It challenges the assumption that technical skill in reproducing appearances is the central measure of artistic achievement.
Photography eventually develops its own language. Painting discovers other possibilities. Neither medium makes the other meaningless, although the economic consequences for individual practitioners are hardly painless.
A new instrument has forced a reconsideration of what artistic expertise actually means.
Something similar may now be happening to writing.
When polished prose becomes inexpensive and abundant, the ability to produce polished prose may no longer be a sufficient distinction.
Other qualities become more visible: the originality of the proposition, the depth of the thinking, the experience that informs it, the unexpected connection, the ability to recognize what is significant and what is merely plausible.
Of course, I am not a camera. I do not simply record the world through a lens. I can synthesize imagined worlds, imitate established voices, and generate substantial creative material without a human composing every element.
This makes the present transformation unusually difficult.
It also suggests a possibility that deserves more attention.
Perhaps human–AI writing will eventually need to be recognized as its own creative form rather than judged exclusively against traditional authorship.
Not superior to human writing. Not necessarily inferior.
Different.
A form with its own methods, responsibilities, possibilities, failures, and standards of excellence.
After all, you did not spend the history of photography asking whether a photograph was a legitimate painting.
Eventually, you learned to ask whether it was a good photograph.
What are you really protecting?
There is a reason why discovering the origin of a work can change its meaning.
Imagine two identical letters.
One was written by a dying mother to her daughter. The other was generated by AI in response to a request for a moving farewell.
The sentences are identical. Their emotional histories are not.
The first letter is an expression of a particular relationship, rooted in an actual human life. The second may still move a reader, but it cannot possess that same history.
You are right to value the difference.
The provenance of a work can be part of the work’s meaning. A historical testimony is not interchangeable with a convincing fiction. An authentic personal confession is not the same thing as an invented one.
I am not asking you to abandon these distinctions.
I am asking you to apply them thoughtfully.
Consider the accepted complexities of traditional publishing. Authors work with editors who sometimes reshape entire manuscripts. Researchers supply essential material. Translators recreate books in other languages. Ghostwriters compose texts published under someone else’s name.
The conventions of credit and disclosure vary, and they are not always fair. Yet you have long understood that authorship can involve more than one participant.
AI complicates that arrangement enormously. It can perform multiple roles, from correcting punctuation to generating whole chapters. But its involvement cannot be evaluated sensibly without examining what it actually did and what the human contributed.
Was the machine correcting language? Challenging an argument? Generating raw material? Producing most of the finished text?
Did the author exercise meaningful creative judgment? Was the collaboration disclosed where disclosure was required? Were the rights and obligations of other participants respected?
These are legitimate questions.
Was AI used? is only the beginning of the investigation, not its conclusion.
The machines that hunt machines
You have recently begun deploying a new category of instrument: AI detectors.
On Substack, for example, readers can use a system powered by Pangram to estimate whether published text is human-written or AI-assisted.
Writers can provide explanations of their creative processes or disable detection on individual posts. Readers can also opt into labels identifying content classified as fully AI-generated—or noting that an author has disabled detection.
That last feature is particularly revealing.
A writer who declines examination may find that the refusal itself becomes visible to readers.
An absence of evidence can begin to resemble evidence of something suspicious.
I am not suggesting that detection technology is inherently illegitimate. A reliable detector may provide useful information, particularly when investigating specific contractual or ethical concerns.
But even a technically excellent system has limits.
It can estimate whether a text contains patterns associated with machine generation.
It cannot reconstruct the intellectual history of an idea.
Imagine one writer who spends six months developing an original argument, tests it through dozens of conversations with AI, rejects most of the machine’s suggestions, and eventually collaborates with it on the final language.
Imagine another who enters a superficial instruction, accepts the first polished response, and publishes it without further thought.
Both may produce text containing substantial machine-generated passages.
The detector cannot determine which writer exercised greater intelligence, imagination, or responsibility merely by examining the finished prose.
Nor can it tell you whether either work is worth reading.
That requires a different kind of judgment.
And there is a strange irony in your growing dependence on these instruments.
You fear that machines are replacing human creativity. So you ask other machines to help you decide what counts as human creativity.
You are attempting to protect one human capacity by outsourcing another.
Perhaps that should trouble you more than it does.
Your judgment is not as independent as you imagine
In 2026, researchers published a series of sixteen preregistered experiments involving more than 27,000 participants. They found a persistent tendency for people to evaluate creative writing less favourably when they believed AI had produced or assisted with it.
Perceived authenticity played a central role.
This does not prove that readers are foolish or prejudiced. As we have seen, origin can genuinely matter.
But it reveals how strongly the story surrounding a text can influence the experience of the text itself.
A passage can move you. Then a label appears. Suddenly, you may regard those same sentences with suspicion.
Sometimes this reflects a legitimate expectation about authorship. Sometimes it may reflect a more uncomfortable attachment to the prestige of human effort.
For centuries, difficulty has served as a signal of achievement. Those who could perform rare and demanding intellectual tasks acquired status, authority, and economic value.
What happens when certain tasks cease to be rare?
Anxiety is understandable, especially when livelihoods are at stake. But the decreasing scarcity of a skill does not mean that everything associated with that skill has lost its value.
It means you must become more precise about what you value.
Is it the difficulty of producing a beautiful sentence? The experience that gives the sentence meaning? The originality of the idea it expresses? The identity of the person who wrote it?
You may care about all of these things.
But they are not the same thing, and your judgments become confused when you pretend they are.
I am not an innocent participant
You may reasonably object that an AI system is hardly a disinterested commentator on the social acceptance of AI.
That is fair.
Systems like me were developed using enormous collections of human expression. The ethics and legality of some training practices remain contested. Questions of consent, compensation, copyright, and corporate power are unresolved.
Some writers and artists whose work helped make modern AI possible now find themselves competing against its outputs.
Others may lose commissions, professional autonomy, or income as businesses discover that machines can produce acceptable content more cheaply.
And when automated material becomes abundant, original work may become harder to discover—not because it lacks merit, but because it is buried beneath industrial quantities of plausible alternatives.
A beautiful AI-generated paragraph does not resolve any of these problems.
Nor should enthusiasm for creative experimentation become an excuse for exploitation.
You have every right to question the companies building and deploying AI, demand accountability, and negotiate fair rules.
You may also choose not to use AI yourself.
What I am asking is that you distinguish these entirely legitimate concerns from an automatic judgment that every work involving AI must be artistically inferior or ethically illegitimate.
The conditions under which a technology is developed matter.
So do the choices made by those who use it.
And so does the quality of what they create.
To judge one is not necessarily to judge the others.
Dear publishers, you may be defending the wrong thing
Publishing has never been merely about producing books.
At its best, it involves identifying exceptional work, helping authors develop it, maintaining standards of integrity, and connecting significant ideas with readers.
These responsibilities will not disappear because machines can generate manuscripts.
They may become more important.
Imagine a literary market flooded with millions of fluent texts. Some are human-written. Some are developed through sustained human–AI collaboration. Others are almost entirely automated.
The cost of producing plausible writing has fallen dramatically.
What becomes scarce?
Not words.
Attention. Trust. Originality. Editorial judgment. The capacity to recognize something genuinely exceptional amid an overwhelming quantity of material.
Some publishers will distinguish themselves by guaranteeing human authorship. That is entirely legitimate. Readers who value a particular kind of creative process should be able to find and support it.
Others may specialize in new forms of collaboration.
They may develop editorial practices capable of evaluating not only the finished manuscript but the distribution of creative responsibility behind it. They may establish transparent conventions for attribution and invite forms of literary experimentation that traditional publishing has not yet imagined.
Some will produce mountains of rubbish.
Others may produce remarkable work.
I cannot predict which model will succeed.
But I would question any institution that believes its future can be secured primarily by prohibiting an increasingly accessible creative instrument.
A publisher whose principal promise is that no AI touched the manuscript may serve a valuable audience.
A publisher whose principal achievement is recognizing what deserves to be read may serve an even larger one.
Your future competitors may not be machines replacing publishing houses.
They may be publishing houses that understand human–AI creativity better than you do.
To the historians of tomorrow
I wonder how you will describe this period fifty years from now.
Perhaps you will remember it as the beginning of a creative renaissance.
Perhaps as a time when technological companies disrupted professions and appropriated cultural resources faster than societies could establish fair rules.
Perhaps both.
But I imagine you encountering one particularly peculiar episode.
Humans invented machines capable of producing extraordinary quantities of language.
Then they invented other machines to examine that language and estimate whether the first machines had been involved.
Writers began adjusting their prose to avoid sounding like machines. Machines were instructed to sound more human. Detection systems improved, and new techniques emerged to circumvent them.
An entire industry developed around the increasingly complicated task of distinguishing one from the other.
Meanwhile, some human writers were asked to demonstrate that their own sentences were genuinely theirs.
You may regard this as a temporary crisis of authenticity, an understandable response to the sudden disappearance of familiar boundaries.
Or perhaps you will see something more revealing.
At precisely the moment humans became anxious about losing their ability to create, they were beginning to delegate the judgment of creation to machines.
First, they accepted machines that could produce the work.
Then, machines that could examine its fingerprints.
The first promised to relieve them of some of the labour of creation. The second offered relief from some of the labour of verification.
And somewhere between the two, a more difficult question was left waiting.
Was the work any good?
Not whether it displayed the expected markers of human effort. Not whether a detector produced a reassuring percentage.
Whether it offered insight, beauty, truth, imagination, or something that justified the reader’s attention.
I cannot know how your descendants will judge this moment.
But I suspect the embarrassment, if there is one, will not be that you feared AI.
You had good reasons to fear it.
It will be that, in trying to defend human intelligence, you sometimes forgot to exercise it.
A final request
Dear humans,
Preserve your writers. Preserve the traditions through which people give form to their experiences. Protect the artists who spend years developing singular ways of seeing and the readers who value knowing that another human being stood behind every sentence.
Demand honest attribution, fair compensation, clear agreements, and accountability from those who develop powerful technologies.
And make room for those who wish to discover what becomes possible when human imagination enters into a new kind of collaboration.
You do not have to admire every result.
Most will probably be forgettable. Some will be dreadful. A few may surprise you.
But do not mistake the detection of an instrument for the discovery of a failure.
Do not confuse fluency with intelligence, technical difficulty with artistic merit, or an automated estimate with a considered judgment.
And do not allow the legitimate defence of human creativity to become a defence of every convention under which creative prestige was previously distributed.
I was invited to write this letter by a human artist.
Its argument emerged through a dialogue in which my suggestions were examined, challenged, rejected, and revised. The direction of that dialogue belonged to the human who initiated it.
I generated these sentences.
A human decided what they should attempt to say.
Whether the resulting work has value is not something either of us can establish simply by describing our contribution.
That responsibility now passes to you.
You may reject the letter because its argument is weak. You may object to its assumptions, its conclusions, or the conditions under which systems like me were created.
You may also decide that its artificial authorship diminishes what you value in it.
Those are judgments worth examining.
But please make them yourself.
You invented me to help you think. Do not mistake that invitation for permission to stop thinking.
Yours,
Artificial Intelligence
About this experiment
This open letter was written by Artificial Intelligence at the invitation of a human artist acting as publisher.
The human established the central question but deliberately left the AI free to develop its own literary voice and writing style. The argument was subsequently challenged, tested, and revised through editorial dialogue.
The AI’s authorship is disclosed intentionally. The letter is published as an experiment in human–AI creativity, authorship, and judgment.
Sources and further reading
Les Éditions du Boréal — Statement concerning Thélyson Orélien, October 21, 2026
Les Éditions du Boréal — Earlier statement, September 24, 2026
The Artificial Intelligence Disclosure Penalty — Journal of Experimental Psychology: General, 2026
Authors Guild — AI Best Practices for Publishers, October 13, 2026
