What does it mean for something to be “written by AI” these days?
Take this article, for example. In fact, take every article on this site.
The ideas are mine. A human mind chose the subject, formed the argument, decided what mattered and supplied the experiences behind it. Then I gave those thoughts to an artificial intelligence and asked it to help me turn them into an essay.
So who wrote it?
The easy answer is that AI did. After all, the machine produced many of the sentences you’re reading. It corrected the grammar, organized the argument, improved the transitions and found cleaner ways to express ideas that were much messier when they first came out of my head.
But that answer feels incomplete.
The smartphone that can beat Magnus Carlsen
A chess program running on an ordinary smartphone can defeat Magnus Carlsen.
That’s an astonishing achievement. One of the greatest chess players who has ever lived can be beaten by a machine small enough to fit in your pocket.
And yet almost nobody wants to watch two computers play chess against each other.
We still watch people.
We watch because the interesting part isn’t merely the quality of the moves. It’s the person making them. It’s the risk, the pressure, the possibility of error and the knowledge that another human being is struggling to find the right move under the same constraints.
The computer may play better chess, but it’s not experiencing the game.
It doesn’t want to win. It doesn’t fear losing. It doesn’t feel the weight of the position or the humiliation of a mistake. There’s nothing at stake for it.
Perhaps writing is beginning to face the same distinction.
AI can already produce remarkably polished prose. It can spell better than most people. It rarely misplaces a comma. It understands structure, rhythm, transitions and the conventions of almost every recognizable form of writing.
But does technical mastery alone make it the writer?
Craft isn’t the same as thought
I’ve been a technical writer for twenty years. I’m considered an expert in my field.
And yet AI can produce a technical document that is, at least on the surface, more polished than something I might write on my own. It can create a logical structure in seconds. It can impose consistency across a document, simplify awkward sentences and organize information with almost mechanical precision.
That’s not easy for me to admit, but pretending otherwise would accomplish nothing.
Still, I regularly find myself correcting it.
Sometimes it misunderstands the technology. Sometimes it gives too much attention to something unimportant while barely mentioning the thing a user actually needs to know. Sometimes every sentence is technically correct, yet the document as a whole is somehow wrong.
The AI knows what a technical document is supposed to look like. It doesn’t necessarily understand the product, the user or the consequences of getting something wrong.
More importantly, it does not care.
That word matters.
It does not care whether the reader understands. It does not care whether the instructions prevent someone from making a costly mistake. It does not care whether the document is honest, useful or misleading.
I care about those things.
The machine can reproduce the shape of concern, but the concern itself must come from somewhere else.
What counts as AI content?
There’s unquestionably such a thing as AI content.
If someone enters a vague instruction such as “write me a blog post about artificial intelligence,” copies the result and publishes it without thought or revision, it seems fair to call that AI generated content.
The machine chose the argument, supplied the observations, created the structure and wrote the prose. The human contributed little more than the request.
But what if the human arrives with the subject, the argument, the analogies, the personal experience and the conclusion?
What if the human rejects paragraphs, corrects misunderstandings, changes the tone, removes ideas that don’t belong and continues revising until the piece expresses exactly what that person wanted to say?
Is that still AI content?
Perhaps. But it’s not the same thing.
We need a better vocabulary for the enormous space between writing every word by hand and pressing a button that says “generate.”
Authorship has never been as simple as asking who physically assembled the sentences. Books have editors. Politicians have speechwriters. Celebrities have ghostwriters. Journalists work with copy editors. Academics receive feedback from colleagues. Writers have always relied on other people to challenge, organize, refine and strengthen their work.
We still recognize the difference between assistance and authorship.
AI complicates that distinction because it can provide all of those forms of assistance at once, almost instantly.
The question of ownership
Maybe authorship isn’t determined by who typed each word.
Maybe it lies in the decisions.
Who decided what the piece was about?
Who supplied the ideas?
Who determined what stayed and what was removed?
Who recognized when a sentence was polished but untrue?
Who accepted responsibility for the final result?
Those questions matter more to me than whether every individual sentence emerged directly from my keyboard.
That doesn’t mean the use of AI should be hidden. Nor does it mean that every person who generates an article should claim complete creative ownership of it. If the machine supplied the ideas, arguments and language, then calling the result entirely your own would be difficult to defend.
But if AI helps a person articulate thoughts that genuinely belong to that person, the situation is much less clear.
The tool has contributed to the expression. It hasn’t necessarily contributed to the thought.
Fiction may be different
This distinction becomes even more complicated with fiction.
A philosophical essay begins with an argument. A technical document begins with information that must be communicated. In both cases, AI can help organize material that already exists.
Fiction asks for something more intimate.
A novel contains the writer’s fears, obsessions, memories, desires and private understanding of other people. Its value isn’t simply that the sentences are grammatically correct or beautifully arranged. It comes from the strange and individual consciousness behind them.
AI can imitate that consciousness. It can create convincing grief, longing, humour or terror. But it’s never grieved, wanted, laughed or been afraid.
Does that matter?
I think it does.
Perhaps readers will eventually decide that it doesn’t. Maybe the emotional reaction produced in the reader is all that matters, regardless of whether anything was felt by the entity that created it.
But I’m not convinced.
A love letter written by someone who loves you means something different from an identical collection of words generated by something incapable of love.
The words may be the same. The meaning is not.
A new kind of editing
Perhaps AI is best understood as an extension of the editing process that has existed since humans first put pen to paper.
The quill became the typewriter. The typewriter became the word processor. The dictionary became spellcheck. The style guide became grammar software. Now the software can reorganize an argument, suggest an analogy, rewrite a paragraph and challenge the assumptions behind an entire piece.
That’s a tremendous change in degree. It may eventually become a change in kind.
But a tool doesn’t erase the person using it.
It changes what that person can do.
The danger isn’t that AI makes writing easier. Tools have always made creative work easier. The danger is that convenience can allow us to stop thinking while preserving the appearance of thought.
Perfectly structured sentences can conceal the absence of an actual idea.
That’s the kind of AI content worth worrying about.
The ghosts in the machine
In a sense, AI is the sum of humanity.
Not literally, of course. It doesn’t contain every human thought or preserve every human voice. But it’s been shaped by an enormous inheritance of human language, knowledge, argument and storytelling.
When AI edits my work and helps shape my thoughts, it can feel as if the ghosts of Hemingway, Orwell, Huxley, Shakespeare and countless other writers are leaning over my shoulder.
Their voices aren’t speaking directly through the machine. What remains are traces of the choices writers have made across centuries. Their rhythms, structures, metaphors, arguments and ways of seeing the world have all contributed to the vast human tradition from which AI learned how writing works.
There’s something comforting about that.
Writers have always learned from other writers. Every book we read leaves something behind in us. Every sentence we write is influenced by sentences written before we were born. AI may simply make that collective inheritance more immediately accessible than it’s ever been.
But there’s also something terrifying about it.
What happens when the accumulated weight of human expression can be summoned instantly? Does it help us discover our individual voices, or does it gradually smooth them away? Does it give more people the ability to express meaningful ideas, or does it flood the world with perfectly constructed sentences that nobody truly needed to say?
I don’t know.
For the record, I suggested adding this section about “the ghosts in the machine” after we had already worked through several drafts and additions. It was the final addition at that stage, and my input had shaped the article throughout.
The whole piece developed through a back-and-forth between me and AI. I supplied ideas, questioned wording and asked for changes. AI offered drafts and suggestions that I could accept, reject or build on. That exchange led to further ideas, additions and revisions.
Even this clarification came out of the same process. I noticed that the earlier wording made it sound as though this section had been my only contribution, explained why that was misleading, and worked with AI to make it clearer.
Make of that what you will.
So who wrote this?
This article began as a collection of my thoughts.
They were unstructured, repetitive and grammatically imperfect. AI helped organize them. It improved the composition. It challenged parts of the argument and found better ways to express others.
But the reason this article exists is that I cared about the question.
I brought twenty years of experience as a writer to it. I recognized what the machine misunderstood. I decided which ideas represented me and which didn’t. I remain responsible for every word I chose to publish.
Was this article written with AI?
Absolutely.
Was it written by AI?
Partly.
Was it written by me?
Yes.
Those answers can all be true at the same time. Perhaps the mistake is assuming that authorship must now belong entirely to one side or the other.
The machine helped shape the language.
The human had something to say.
Want to pick better books?
Tell us what you love, what you dislike and what you’re in the mood for. We’ll find books matched to your taste and explain why they fit.
Find my next book