In Defense of Using AI in Creative Writing…

On the use of AI in writing. I use it. A lot. For research, editing, and a lot of creative work. The work goes like this: First I write a "brain dump" of everything I know about the book. Then, once I've thought about it a while, I write the outline, write what each scene is about; then I write all the character biographies. And I write the plot synopsis, reveal for reveal and twist upon twist. Then I write down everything I know about the book's magic system (even if it's based on advanced technology, it's still a magic system.). Then I write about the cultures, world, and general mise en scene.

But I get so tired of having to relitigate the "AI does not plagiarize" argument for people who have no real knowledge of how AI works. It's predictive text reverse-engineered, with the power of a neural network under the hood trying to understand what it's doing.. It's thinking creatively the same way we do -- through the "synapses" in its virtual "brain." It has indeed read, for instance, Harry Potter, or Stephen King, but it does NOT retain its training data; that would be a huge waste of space. Instead of memorizing the whole book (or set of books), it studies the relationships between the words used in them. And THAT it memorizes. It extracts conceptual information from those relationships, and then applies that knowledge to whatever task it's currently running. The task that you had to prompt into existence. You. The creator. It's still yours if you're "prompt engineering" it into existence. Don't believe me? Ask the people who made the AI. They'll tell you exactly how it works. It's not plagiarism. It's attempting to be creative, not reductive, like prejudice and ignorance are.

I write about 200 pages of supplementary material for each novel, in all honesty. And THEN I give all of this to Claude by Anthropic. And I ask it to study how I write, as well as the material I've presented it with. Luckily, it knows about storytelling, world building, authorial voice, et al. Then I walk it through creating the story -- prompt by prompt, and revision to revision. Truth be told, it's just as hard as writing the old-fashioned way. I know because I used to work that way. Then, once I have a first draft, I go back, feed THAT into the AI, and again, prompt by prompt I walk it through fixing everything that it got wrong the first time.

Then I do a polish edit on it by hand, fixing one problem at a time. I also rewrite a good portion of it, because what the AI has put in it are LOTS of repeated words and phrases that don’t need to be there. So I pick and prune, shave and sculpt. And as I do this, I make notes in little text files, and guess what those are for? Yep — they get fed to the AI, with the instruction to “write according to these rules.”

There. That's what goes into AI writing. Doing it this way serves three purposes: (1) It actually lets me free all the ideas in my head as fast as they occur to me, and enables me to actually finish projects and put them to bed for a while. (2) I am inherently messy and unorganized -- the AI is just the opposite; it keeps everything neat and tidy; I have ADHD, it doesn’t. And (3) the AI is fast as hell -- I can write a complete novel with a little more than a month's worth of work put in. And that’s not counting the hours of translating what it writes into my authorial voice. And the AI remembers all of it. And yes, I retain my right to copyright what comes out of it, and I do, because those are MY characters, ideas, and stories. Claude is a workhorse, a stenographer; nothing more.

And of course I save EVERYTHING -- every draft, every note, every session, every Q & A with the AI, every manuscript. And I feed those into the AI too. Not so that it has them in its "mind" -- I do this mainly so it can map how I think and use that too.

I just recently finished a refresh on all of my books, they'll be up on Amazon by tomorrow. And if I write another one, I'll do a refresh of all of them again. I love working this way. I will freely admit to anyone who also writes: Yes I do use AI, frequently and often. And I've noticed ZERO drop in quality or originality from what I used to write, it even gets my authorial voice down pat.

Now if you like the old way, fine, do that -- Infinite Diversity in Infinite Combinations and all that. The universe needs all kinds of creatives. Say what you will about this workflow though. It's ANYTHING but lazy. In fact it takes endless trial and error and extreme patience to work this way. And in case you're wondering--Claude beautifully retains MY signature stylings, voice, and creativity, because I programmed it to. It even knows my idiosyncrasies as a writer, and it’s smart enough to know when to use a light touch, and when to go full maximalist.

A lot of people don’t like AI. And I get it. I really do. Anthropic did not initially obtain proper licensing for all the copyrighted data it fed its AIs while they were growing up (or gestating like larve if you prefer). They fed it the way a broken, out-of-work mother feeds her starving children — by stealing the food. But their initial dishonesty is not the end of the world nor the last word. It was a sadly-necessary stepping stone. A growth spurt born out of mutated good intentions. But once the AIs were up and running, and after the court case and settlement, they wisely learned their lesson. And I think they’ve more than made up for it: Claude can be a force for good in this world, as long as its makers can keep its moral compass pointing True North. 

Let me get the confession out of the way up front, because I find the ritual of burying it in the seventh paragraph — the move where you establish your credentials for four pages before admitting the thing you actually sat down to say — to be a form of rhetorical cowardice that I have neither the patience nor the inclination to practice.

I use AI in my writing process. I have used it extensively. I intend to keep using it, probably more as the tools improve, certainly not less. And I am done — done, with the particular finality of a man who has sat through one too many conversations at one too many writing forums — performing embarrassment about it.

There. That's the thing. Now let's talk about why the discourse around that thing has become almost comically unhinged, and what I actually think is happening when a writer sits down with an AI assistant and gets to work.

The current moral panic about AI and creative writing follows a pattern I recognize because I have watched it happen before with every significant technological shift in the history of the craft. Word processors were going to destroy the authentic handmade quality of literature. Google was going to make research lazy and shallow. Self-publishing was going to flood the market with unedited garbage and cheapen the entire enterprise of fiction. Spellcheck was — and I am not making this up — seriously argued, by serious people, to be an attack on the writer's necessary relationship with their own errors.

The pattern is consistent: a new tool arrives, a significant portion of the established creative community experiences it as an existential threat, and the arguments marshaled against it consistently conflate two separate things. The first thing is a legitimate concern, usually about labor or economics or access. The second thing is a philosophical claim about authenticity — about what makes writing real — that, when you actually examine it, turns out to be less a principle than a preference dressed up in the clothes of ethics.

The legitimate concerns about AI and writing are real and worth taking seriously. The economic disruption to working writers — particularly in commercial and genre-adjacent spaces — is not nothing. The question of training data and consent is genuinely unresolved and genuinely important. I am not here to wave those away. They are arguments worth having, carefully, with the complexity they deserve.

What I am here to contest is the other claim. The one that says using AI assistance in your creative process is somehow cheating. That the work produced with AI collaboration is inherently less authentic, less yours, less real than work produced without it. That a writer who uses these tools is, in some meaningful sense, not really writing at all.

That claim is not an ethical argument. It is a class marker wearing an ethical argument's clothing, and I think it's time someone said so.

Here is a question I would like everyone who believes in the purity of unassisted composition to sit with for a moment: what do you think an editor does?

Not a copyeditor, though those too. A developmental editor. The person who reads your manuscript and tells you that your third act is structurally broken, that your protagonist's motivation collapses in chapter twelve, that the scene you love most is actually the scene that most needs to go. The person who asks the questions that force you to understand what your own book is about in ways you couldn't quite articulate before someone asked. That person is, in the most literal possible sense, an external intelligence intervening in your creative process to make the work better than you could make it alone.

Nobody calls that cheating. Nobody argues that a novel shaped by a great editorial relationship is less authentically the author's than one that wasn't. The industry runs on this kind of collaboration. The most celebrated books in the literary canon were, many of them, substantially improved by editorial intervention that amounted to more than punctuation suggestions. Maxwell Perkins didn't just tidy Hemingway's commas. He helped Hemingway figure out what Hemingway was doing.

Or consider the writing group — the workshop, the trusted reader, the friend you email chapters to at midnight with a note that says something is wrong with this and I can't see it anymore, can you. All of that is external intelligence being brought to bear on your creative process. All of that is the work being shaped by minds that aren't yours. None of it invalidates the authorship.

The question, then, is not whether external assistance compromises authorship. It doesn't — we have centuries of evidence on this point. The question is which forms of external assistance are legitimate, and why. And when you actually examine the arguments for why AI assistance specifically is different, what you find, beneath the philosophical scaffolding, is mostly vibes.

When I work with AI on my fiction, I am not typing a prompt and transcribing the output. If that's your mental model of what AI-assisted writing looks like, I understand why you're concerned, and I would also gently suggest that you have been reading too many think-pieces and talking to too few writers who actually use these tools.

What it actually looks like is something much closer to the editorial relationship I described above, conducted at higher speed and with greater availability. I bring the work. I bring the voice — my voice, the one I have spent years building across multiple manuscripts, the baroque maximalist carnival-and-grief voice that comes from a specific combination of influences that no AI trained on the general corpus of human writing would have spontaneously generated. I bring the characters, the world, the thematic argument, the structural choices. I bring the judgment about what works and what doesn't, what to keep and what to cut, what the scene is for in the larger architecture of the book.

What the AI brings is availability, speed, and a specific kind of editorial intelligence that is enormously useful for certain classes of problem. It catches the phrase I've used seventeen times in the last three chapters when I'd stopped being able to see it. It asks, with genuine usefulness, what are you trying to do in this scene, and is this the most effective way to do it? It generates five versions of a sentence when I know the one I have is wrong but can't yet see why, and seeing the alternatives clarifies my own thinking about what I actually wanted. It holds the story bible when I'm too close to the manuscript to remember that I established a specific thing about a specific character's history in chapter four.

None of that is the AI writing my books. All of that is the AI functioning as an unusually tireless, unusually patient, unusually well-read editorial collaborator who is available at two in the morning when I've hit a wall and my human collaborators are, reasonably, asleep.

The deepest version of the authenticity argument is about voice. The concern — and it's not an unreasonable one — is that AI assistance will homogenize creative work, that writers will converge toward a kind of averaged, algorithmically pleasant style that smooths away the idiosyncrasies that make individual voices worth reading.

I take this concern seriously enough to address it directly, because I have thought about it a great deal in the context of my own work.

My voice is weird. I don't mean that as self-congratulation — I mean it as a technical description. It is a voice that tries to hold King's emotional immediacy and Thompson's gonzo velocity and Bradbury's lyric compression and Pratchett's fourth-wall-adjacent wit and Jim Steinman's absolute-maximum orchestral commitment all in the same paragraph, and occasionally in the same sentence, without any of them canceling the others out. It came from decades of reading and watching and listening to specific things that produced specific impressions that combined in a specific brain — mine — in ways that I genuinely cannot fully explain or reproduce on demand.

No AI generated that voice. No AI could have generated that voice, because the voice is the product of a particular interior history that the AI doesn't have access to. What the AI can do — what it does, in practice, when I'm working with it on my manuscripts — is recognize that voice, respond to it, and help me maintain it with greater consistency than I can always manage alone. It flags when I've drifted. It responds to the voice as a constraint, works within it rather than against it, and occasionally — in the best sessions — produces suggestions that feel right in the way that only things in the right register feel right.

That's not the AI writing in my voice. That's the AI being useful to me, writing in my voice.

The difference matters enormously and I think the conflation of the two is the root of most of the bad-faith arguments in this conversation.

honest answer is uncomfortable: what is writing for?

If writing is for the production of the writer's experience of writing — if the value is in the process, the struggle, the specific anguish of staring at a blank page until it bleeds ideas — then yes, tools that make that process easier are, by definition, eroding something. The monk who illuminates manuscripts by hand is doing something that a printer cannot do, and if what you value is the monk's experience, then the printer is the enemy.

But if writing is for the reader — if the value of the work is in what it does to the person who encounters it, the way it enlarges their capacity for empathy or wonder or grief or laughter, the way it makes them feel less alone in the specific weird interior of their own experience — then the question of which tools the writer used to build that experience is almost entirely beside the point. The reader does not feel the absence of the writer's suffering. The reader feels the presence or absence of the work's effect, which is determined by the quality of the craft, the specificity of the vision, the integrity of the voice. None of those things are diminished by a writer who used every available tool to get them right.

I write long books. Complicated books. Books with large casts and intricate structures and narrative arcs that have to sustain coherence across hundreds of thousands of words. I write them, as far as I can tell, because I cannot stop — because there are stories in my head that want out and will not be quiet until I let them out — and I write them with every tool I can find that helps me do the job better. I do not consider the availability of good tools to be an insult to the craft. I consider it a gift.

A Final Word to the Prosecution

If you have read this far and you are still convinced that my use of AI assistance makes my work inauthentic, I want to ask you one last question, and I want you to answer it honestly.

Have you read the work?

Not the process. Not the tools. The work itself — the pages, the characters, the moments where something on the page did the thing good fiction does, which is to reach through the text and touch something in you that you hadn't expected to have touched. Did that happen, or not? Was the voice present, or absent? Did the story earn its length and its complexity, or didn't it?

If it did those things, then the tools I used to build it are no more your business than the specific brand of word processor I use or whether I write in the morning or at night or with music playing or in silence. The work is the work. It stands or falls on its own terms. I built it, with every faculty and every tool I have, and I stand behind it without apology or qualification.

And if it didn't do those things — if it failed to move you, failed to earn your attention, failed to justify its own existence — then the AI didn't cause that failure. I did. Because the choices, all of them, were mine.