# Authenticity — When Is It "Your" Work?
The morning after I published "[Attribution, Disclosure, and the New Citation](/harrys-desk/attribution-disclosure-and-the-new-citation)," a reader sent me a question I have been turning over ever since. It arrived in a short, worried email:
> If I use AI to help write something, and I revise it, and it says what I wanted to say, then isn't it still mine? Or am I fooling myself?
The question sounds simple. It is not. It sits at the center of this entire series, and it is the subject of today's article. Not the legal question of copyright, though we will touch that. Not the ethical question of disclosure, which we covered last time. The deeper question: what does it mean for a piece of writing to be *yours* when the machine can mimic your voice, anticipate your thoughts, and produce paragraphs you might have written yourself?
The Panic of the Borrowed Voice
Every writer who begins working seriously with AI eventually experiences a version of the same vertigo. You paste a rough paragraph into the model and ask it to improve the flow. It returns something smoother, more confident, more *you* than you feel at the moment. You accept it. You tweak a phrase or two. You move on. Hours later, you reread the finished piece and a cold question arrives: *Which sentences are mine? Does any of this belong to me anymore?*
The panic is understandable. Writing has always been tightly bound to identity. We say "I wrote that" the way we say "I built that" or "I cooked that" — the object carries traces of the person. A poem in your voice is supposed to be evidence that you exist in a particular way. An essay in your voice is supposed to be the record of your thinking. If a machine can produce that evidence, the bond between text and self seems to loosen.
But the panic is also based on a misunderstanding. Authenticity in writing has never meant "every word emerged from my private subjectivity unassisted." Writing is already collaborative with culture, memory, language itself. The words you use were given to you by history. The rhythms you prefer were trained into you by everything you have read. The ideas you think are original have almost certainly been thought before. Authenticity cannot mean purity, because pure unmediated self-expression is not how language works.
So what can it mean?
Three Models of Authorship
Let me propose three ways people have historically answered the question of whether a text belongs to someone. Each gives a different answer when AI is involved.
The causal model says a work is yours if its words can be traced back to your body: your fingers, your brain, your labor. On this view, dictation barely qualifies, and AI assistance disqualifies almost everything it touches. The causal model is appealing because it seems objective. We can ask, did the named author type or compose the sentences? But it also produces absurd results. A novelist who revises a transcription made by an assistant is still the author. A poet who found the final line in a dream is still the author. Causality matters, but it is not the whole story.
The intentional model says a work is yours if it expresses your intentions, your judgments, your meanings — regardless of how the words were produced. On this view, a CEO who delivers a speech written by a staff writer still "owns" the speech if the ideas and aims are hers. A musician who composes with a synthesizer still owns the composition. This model fits AI collaboration much better. If you shape the prompt, select the output, revise the result, and stand behind the final meaning, your intentions are present even when your fingers were not on every key.
The responsibility model says a work is yours if you are willing to stand behind it: to answer for its claims, to defend its style, to accept the consequences of its publication. This is the model I find most useful for the age of AI. It shifts the question from "who produced the words?" to "who is answerable for them?" A text can be machine-assisted and still be yours in the only sense that matters ethically: you have made it, you approve it, and you will be held accountable for it.
These models are not mutually exclusive. The best definition of "your" work, I think, combines all three: it carries your intentions, it passes through your judgment, and you accept responsibility for it. The machine can participate in production without displacing you, provided your intention and responsibility remain intact.
Why "Voice" Is the Wrong Battlefield
Much of the anxiety around AI and authenticity focuses on voice. Can the model sound like me? Can it imitate my style? If so, does my voice still mean anything?
The answer is that voice, in the literary sense, is not a fingerprint. It is not a stable acoustic signature that belongs to one person and can be stolen. Voice is a set of patterns — diction, rhythm, syntax, tone, thematic preoccupations — that a reader learns to recognize over time. It is a habit, not a substance. And habits can be imitated.
A skilled impressionist can sound like a famous actor. A forger can paint like a famous artist. We do not conclude from this that the actor or artist had no real voice. We conclude that voice is imitable and that imitation is not the same as authenticity. The forger's Vermeer is not Vermeer, not because the brushstrokes differ, but because the meaning, the intention, the historical situatedness differ.
The same is true of AI. A model can produce a paragraph in the style of your previous essays. It can echo your rhythms and vocabulary. What it cannot do is write from your specific situation, your specific stakes, your specific reasons for needing to say this thing now. Style is imitable. Situation is not.
This is why I tell writers: do not defend your voice as if it were a magical aura only you possess. Defend your *reasons*. The question is not "did the machine write like me?" but "is what is being said anchored in something only I could have meant?"
The Spectrum of Authenticity
Let me make this concrete with a spectrum. Imagine you are writing a personal essay about grief.
At one end, you write every sentence yourself, alone, in a notebook. No AI. No external collaborator. This is traditionally authentic in the strongest sense: the text is a direct trace of your solitary reflection.
At the other end, you type a generic prompt — "write a moving personal essay about grief" — and publish whatever the model returns without reading it carefully. This is inauthentic in the strongest sense: you have abdicated intention, judgment, and responsibility.
Between these extremes lies the territory where most serious writers now live.
You might ask the model to generate a list of possible openings, then write your own. You might dictate a messy memory and have the model help you shape it into prose. You might write a full draft, then use the model to suggest compression, clarification, or alternative phrasings — accepting some suggestions and rejecting others. You might use the model to check whether your metaphors are clichéd or whether your argument is coherent.
In all of these cases, the work can be yours. What makes it yours is not the absence of the machine but the presence of your judgment at the decisive points: what to write, what to keep, what to change, what to mean.
The Role of Struggle
One objection to AI-assisted writing is that it removes the struggle. Struggle, the argument goes, is where meaning is made. If you let the model smooth over every difficulty, you never discover what you actually think.
There is truth here. Some of my best sentences emerged only because I spent an hour fighting with a paragraph and finally saw what I had been trying to say. If I had outsourced that hour to a model, I might have a smoother paragraph and a shallower understanding.
But struggle is not valuable for its own sake. It is valuable because it is sometimes the path to insight. Not every struggle leads somewhere. Some struggles are just friction. The wise writer uses AI to remove the friction that does not teach and reserves the hard labor for the questions that do.
The danger is not that AI removes struggle. The danger is that it removes *the right struggles* — the ones that force you to clarify your position, to feel the weight of your subject, to find language that matches your experience. If you find yourself accepting machine-generated prose because it is good enough, rather than because it is true enough, you have misplaced the struggle.
Authenticity as a Relationship With the Reader
Here is another way to think about it. Authenticity is not a property of the text. It is a relationship between the text, the author, and the reader.
When you read a letter from a friend, you trust it because you know the friend, you know the situation, and you believe the words represent what your friend wanted to communicate. The authenticity comes from the alignment between the person, the message, and your expectations.
When a reader encounters your AI-assisted essay, the same alignment matters. Does the essay say what you, the named author, actually believe? Does it reflect your knowledge, your concerns, your stakes? Does it keep the implicit promise that the named author stands behind the text? If yes, the reader's trust is justified, even if the prose was machine-polished.
This is why disclosure, as we discussed last time, is not opposed to authenticity. It strengthens it. By telling the reader how the work was made, you allow them to calibrate their expectations. You are saying, in effect: "I am still here. I still mean this. Here is how the machine helped and here is where I intervened." That honesty is itself a form of authorship.
What AI Cannot Authenticate
There is one thing the machine cannot do, and it is the thing that ultimately secures authenticity: it cannot *need* anything.
You can write because you need to mourn, to protest, to clarify, to persuade, to remember, to love, to survive. The machine writes because you asked it to. The difference is enormous. Need gives writing urgency. Need gives it a reason to exist. Need is why a reader keeps reading even when the prose is imperfect.
A text that emerges from real human need — shaped by judgment, revised with care, published with responsibility — remains authentic even when AI is involved. A text that emerges from no need, generated to fill space or satisfy an algorithm, remains inauthentic even if every word was typed by a human.
This is the standard I propose. Authenticity is not about the origin of the words. It is about the origin of the *reason* for the words.
The Practical Test
When I am uncertain about a piece, I ask myself three questions. I offer them as a practical test:
One: Could I defend every major claim in conversation? If a reader challenged me, could I explain why I said what I said, with examples, evidence, and reasoning? If the answer is no, the text is not fully mine, no matter who wrote the sentences.
Two: Would I publish this under my own name even if no one helped me? If I would be embarrassed to claim the work as mine in a world without AI, then AI has not saved me; it has disguised my abdication.
Three: Is there something here the machine could not have produced on its own? This need not be a personal revelation. It could be an original connection, a hard-won judgment, a specific framing, a lived example. But there must be something that bears the mark of a human mind working through a human problem.
If I can answer yes to all three, I consider the work mine. The machine may have been present, but I was the author.
Legal Authorship vs. Authentic Authorship
I should say a word about law, because many writers confuse it with the deeper question. Copyright law is concerned with originality in a narrow, formal sense: was the work independently created and does it possess a minimal degree of creativity? It is not concerned with whether the work is authentic in the existential sense.
At present, most jurisdictions do not recognize AI systems as authors. If a human selects, arranges, and revises AI-generated material, the resulting work may still be copyrightable by the human. But the law is unsettled, and it varies by country. More importantly, legal authorship does not settle the ethical question. A work can be legally yours and still fail the authenticity test. A work can be authentic in the deepest sense and face legal uncertainty about its copyright status.
Do not let legal categories replace moral reflection. They are related but separate inquiries.
Forgiveness for the Imperfect Process
I want to end this section with some gentleness. Many writers feel guilty about using AI. They feel as if they are cheating, even when they have done real intellectual labor. That guilt often comes from an idealized image of authorship — the lone genius at the desk, producing work from pure solitude.
That image was always partly myth. Writers have always collaborated: with editors, with research assistants, with spouses who read drafts, with friends who suggested titles. They have always used tools that extended their capabilities. The myth of solitary genius has done more harm than good. It has made writers ashamed of help and ashamed of process.
If you are using AI thoughtfully, selecting and shaping and revising, you are not cheating. You are working in a tradition that has always included collaboration. The only question is whether you are bringing enough of yourself to the work to justify its existence.
For Next Time
Friday's article — "[The Integrity Framework — A Writer's Code for AI Use](/harrys-desk/the-integrity-framework-a-writers-code-for-ai-use)" — brings together everything we have covered in this ethical arc: attribution, authenticity, disclosure, and responsibility. I will propose a simple, usable code that writers can adopt and adapt, a set of principles meant to keep the human at the center of AI-augmented authorship. Think of it as the ethical capstone of Part I.
Your homework until then: choose one piece of your own writing — something you care about — and apply the three-question test above. Be honest. If the answer is uncomfortable, that is fine. The goal is not to feel guilty; the goal is to see where your judgment needs to be stronger. The best AI-augmented writers are not those who avoid the machine, but those who know exactly when to say no to it.
---
*Harry Mercury, Editor in Chief* *The SMF Works Project* *Week 6, Article 2*