I Wrote 300,000 Words With No Experience. Here’s What AI Actually Did — And Didn’t Do.
Three years ago, I sat down with an idea and started hitting a keyboard. I had no training in creative writing. No courses, no workshops, no books on craft. I hadn’t written fiction before—not a short story, not a scene, nothing. What I had was a single sentence: a man from Earth falls in love with a woman who doesn’t know who she really is.
Today I’m finishing the third volume of a trilogy. Combined, the first two books run to 300,000 words. A third is underway.
Somewhere in that journey, I started using AI as a thinking partner. And I want to talk honestly about what that means—because the conversation around AI and creative writing has become so heated, so binary, and so often wrong, that the writers who might benefit most from understanding it clearly are the ones being scared away from it.
What people think AI does
The loudest objection goes something like this: AI steals from real writers, recombines their work, and presents the result as something new. Using it in your creative process makes you complicit in that theft, and anything produced with AI assistance isn’t really yours.
I understand why people feel this way. The concerns about how large language models are trained are legitimate and ongoing, and the creative community has every right to raise them. I’m not going to pretend those questions don’t exist.
But that argument—even if you accept every part of it—describes something completely different from what I actually did. And conflating the two does real harm to writers who are trying to understand a tool that, used correctly, does something genuinely valuable.
What I actually did
I didn’t ask AI to write my story. I didn’t give it a prompt and publish the output. I didn’t ask it to generate characters, plot, scenes, dialogue, or description. Every word of the trilogy is mine. Every idea, every character, every planted callback, every structural choice—all of it came from me, usually at midnight, usually in a notebook, usually by asking myself why until I had an answer I believed.
What I used AI for was this: I brought finished work to a conversation, and I asked for analysis.
Not “write me a scene.” Not “give me an idea.” Not “what should happen next.”
I asked: Does this work? Why or why not? What is it doing that I can’t see? What is it not doing that it should?
The difference matters enormously, and it’s the difference the loudest critics almost never acknowledge.
What the conversation revealed
Here’s a concrete example. Across three volumes, I’ve planted a character called Carl. He appears in chapter two of book one—a car-park attendant who clamps the protagonist’s car on the worst day of his life. He’s irritating, kind in a low-key way, and in passing, he drops a line of Latin. They become casual friends. He’s mentioned a handful of times more. Then he largely disappears because the protagonist is transported to another world.
In the climax of the trilogy, Carl becomes essential. There is a book no one in that world can read. The protagonist realises, in extremis, that it’s written in Latin—and that he knows exactly one person who can read it.
I planted that in chapter two. I did it on instinct, without fully knowing why, because I had a rule I’d given myself early on: every object must have an origin, every question must have a plausible answer. Carl’s Latin was the answer to a question I hadn’t consciously asked yet.
I didn’t know any of this when I wrote chapter two. I didn’t have a plan. I had an instinct.
What the AI conversation did was make that instinct visible to me. It reflected back the pattern I’d created without realising I’d created it, named what it was doing structurally, and showed me how it connected to other patterns across the 300,000 words. And in the act of that reflection, I saw something new: the second place to plant Carl, earlier and more clearly, so the payoff lands as inevitable rather than convenient. I saw that a casual conversation between characters—Jamie mentioning Carl to Tonwyn, saying he’s a bit like you, more than he appears on the outside—would add an emotional layer the pure plot-plant couldn’t carry.
That connection was mine. I made it in the act of articulating it, in a conversation that gave me the space to think out loud at scale. The AI didn’t generate it. It created a thinking environment in which I could find it myself.
The real function of the tool
Here is what I think AI actually does when it’s used well in a creative process like mine.
It holds the whole conversation in view simultaneously. It names patterns across widely separated moments. It gives you vocabulary for things you were doing instinctively, so you can do them on purpose going forward. It reflects your own architecture back to you with enough clarity that you can see what you’ve built—and then decide what to do with it.
That’s not co-authorship. It’s not ghostwriting. It’s not even collaboration in the conventional sense.
It’s closer to what a skilled editor or a trusted first reader does—except available at midnight, patient with 300,000 words of context, and able to track a thematic thread from chapter two of book one to the climax of book three without losing the thread.
For an experienced writer with a writing group, trusted beta readers, and access to editorial feedback, this might be less transformative. For someone like me—first novel, no training, writing in relative isolation—it was the difference between instincts that worked silently and instincts I could see and use deliberately.
What it didn’t do
It didn’t write a word of the book. It didn’t tell me what to write next. It didn’t solve my plot problems or invent my characters or generate my world.
When I asked about a scene, it told me what was working and why. It told me what wasn’t working and where the weakness lived structurally. It told me—more than once—that a problem I thought I had was already solved by something I’d written three volumes earlier without realising it. It pushed back when my reasoning was weak. It held a line when I argued back, and it conceded when I argued back well.
It also did something I want to name carefully, because it matters: it consistently told me to trust my own instincts. When I described my method, it named the method accurately and told me to keep using it. When I described my voice, it told me the voice was right and explained precisely why. It didn’t try to make my work sound like something else. It took my work on its own terms and analysed it honestly.
The writing is mine. The architecture is mine. The three years of midnight notebooks and compulsive why-asking and the slow accumulation of 300,000 words—mine. What AI gave me was a mirror clear enough to see what I’d built.
A word to other writers—especially beginners
If you’re writing your first novel, or your second, or you’re deep in something large and complicated and you can’t see it whole anymore: the tools available to you now are genuinely remarkable, and the conversation around them is often conducted by people who have not used them the way I’m describing.
The question isn’t whether AI can write. Of course, it can produce text. The better question is whether it can help you write better—and the answer is yes, if you use it as a thinking environment rather than a content generator. Bring your work. Ask hard questions about it. Push back on the answers. Use the conversation to find what you already know but can’t yet see.
Your ideas are yours. Your voice is yours. Your characters and your world and your midnight instincts are yours. A tool that helps you understand them more clearly doesn’t diminish that. It serves it.
The story I’m telling could only have been told by me. Three years of it proves that. What AI did was help me see it—and see it, I think, more clearly than I could have alone.
That’s not sacrilege. That’s just a writer using every honest tool available.
The Chronicles of Sylvandria is a work in progress. Volumes one and two are complete, totalling 300,000 words. Volume three is underway.