How I Write With AI (And Why It’s Still My Voice)
I don’t sit alone in a cabin with a typewriter. I don’t bleed onto the page, one key at a time, chain-smoking and glaring at a blank sheet like it’s a personal enemy.
I write in a chat box.
On my good days, that still feels faintly scandalous like I’ve snuck into some forbidden wing of the library where the books write back. On my bad days, I hear the chorus: “That’s not real writing.”
Here’s the thing, though. Every sentence you’re reading started as my idea. My questions. My weird little obsessions. AI is in the room with me, sure, but it’s not the author. I am.
The Collaboration
The basic rhythm goes like this. I open ChatGPT (just Chat to me—easier to say and type) and say something like:
“I want to write about how I actually use AI to write. I want to talk about disability, access, the myth of ‘AI detectors,’ and the fact that these are still my ideas. Let’s aim for around a thousand words.”
Then, it responds with a draft. Sometimes rough, sometimes surprisingly sharp. From there, the real writing begins.
I cut whole paragraphs. I reframe sections. I add stories from my life; I strip out anything that sounds too slick or generic. I tweak phrasing until it sounds like me: the same cadences that show up when I talk to my wife in the kitchen or leave a voice memo for a friend.
What AI gives me is motion. It breaks the inertia of the blank page. Instead of staring at nothing, I get to argue with something. I can say:
“This part is close. Push it more.”
“This sounds like a brochure; try again, but more essay, less pitch.”
“You skipped the hard part. Go into that.”
I’m not outsourcing my voice. I’m drafting with a tool that happens to write back fast.make all the difference.
These Are my Ideas
There’s a real fear, understandable, about AI turning everything into the same slurry of fake insight and half-remembered tweets. I share that worry. I’ve read plenty of AI-generated sludge. I don’t want to add to it.
So, here’s the line I draw for myself.
I don’t ask AI what I should think. I don’t say, “Write me a hot take about AI and creativity,” then hit publish. I come in with my own questions and commitments.
I care about disability and access.
I care about honesty in how we talk about these tools.
I care about attention and how we spend our limited time.
I am the one who brings those values to the conversation.
When I say “write about how it’s not really possible to reliably detect AI writing,” that’s not a prompt I pulled from a growth-hacking newsletter. That’s coming from reading, from paying attention to how people actually use these tools, from my own discomfort when someone confidently claims, “I can always tell.”
The model offers structure, transitions, different ways of saying what I already believe. Sometimes it suggests an angle I hadn’t quite seen. But the spark? The tension? The “why bother saying this at all?”—that’s mine.
Disability, Access, and why This Matters
I’m legally blind. I live with progressive MS. Fatigue isn’t just “tired after a long day”; it’s a hard limit. I don’t get to pretend my energy is infinite.
Old-school romantic writing culture—the long nights, the heroic suffering, the idea that real art requires maximum difficulty—has never been less appealing to me. I’m not interested in performing struggle for its own sake.
AI lets me spend less of my tiny daily energy budget on wrestling formatting, or retyping whole paragraphs from scratch, or starting from nothing again and again. It lets me focus on the parts that actually feel like writing.
Finding the emotional core of what I want to say.
Choosing which details to keep and which to kill.
Deciding what I’m willing to stand behind publicly.
For disabled writers, tools that lower friction aren’t “cheating.” They’re access. They’re ramps instead of stairs. No one looks at a ramp and says, “You didn’t really enter the building.”
I still have to decide where to go once I’m inside.
The Ghost of the Typewriter
We carry this inherited picture of what “real writing” looks like: one person, one machine, one draft at a time. Typewriters, legal pads, red pens. The lone genius hunched over a desk, protecting the purity of their process.
It’s a gorgeous image. It’s also a historical moment, not a law of nature.
Before typewriters, there were quills and scribes and dictation. Before private rooms, there were noisy tables and shared manuscripts. Writing has always been a technology stack: tools, habits, constraints.
Now, the tool happens to talk back.
When I chat with an AI, I’m not betraying some sacred contract with the ancestors. I’m updating the stack. The process looks different, but the questions I wrestle with are weirdly timeless:
What am I trying to say?
Who am I saying it to?
What am I willing to put my name on?
The myth isn’t that AI is neutral or harmless—it isn’t. The myth is that there was ever a single correct way to make sentences.
“I Can Always Tell if AI Wrote This”
You’ve probably heard that one. Maybe even said it.
There was a brief era when AI writing really was easy to spot. Those early, overconfident essays loved certain words and rhythms. Everything wanted to delve into topics. Every paragraph promised to explore something “in today’s fast-paced world.” You could feel the template showing through the paint.
We’re not in that era anymore.
Models have improved. Prompts have improved. And, just as important, more humans are in the loop—revising, cutting, reshaping. A piece might start out with that slightly plasticky voice and end up sounding like an actual person after enough passes. At that point, what exactly are you “detecting”?
Here’s the awkward truth: at scale, reliable detection just doesn’t hold up. AI detectors get things wrong—often. Short texts, mixed human/AI edits, translations, heavily revised drafts…the lines blur fast. Even when tools claim a confidence score, they’re still making probabilistic guesses, not reading some secret watermark of “realness.”
Then, there’s the basic messiness of writing. Ghostwriters exist. Heavy-handed editors exist. Teachers rewrite student drafts. Friends text suggested lines for wedding toasts. Raymond Carver’s famously spare stories didn’t emerge untouched from his typewriter; Gordon Lish carved into them, rearranged, cut, reshaped. You can argue about whether those edits helped or hurt, but you can’t pretend there was a single, untouched “pure” Carver to measure everything against.
Oh, and then there’s the em dash tell/lie. There are ten in this essay—I checked. Some put in (or removed) by me and some by Chat. Does it really matter which?
So, imagine some detector scanning a hybrid piece like this and declaring “90% AI-generated.” Does that erase the hours I spent editing? The memories I decided to share and the ones I didn’t? The values that shaped what I left in?
Detection fantasies appeal because they promise an easy way to sort “authentic” from “fake.” But writing has never been that tidy.
I’d rather ask, “Are you being honest about your process?” or “Are you taking responsibility for the words you release?” or “Are you treating other pe,ople’s time and trust with respect?”
If the answer is yes, I don’t care how many drafts passed through a language model along the way.
Still a Person at the End of the Line
The most important part of my process doesn’t happen in the chat box. It happens when I sit there, rereading, and decide what stays.
I’m the one who chooses to keep the image of the typewriter and cut some other clever metaphor. I’m the one who decides to bring in disability and access instead of pretending my use of AI is purely a fun productivity hack. I’m the one who hits publish—or doesn’t.
AI accelerates the mechanics. It speeds up the path from “vague feeling” to “readable paragraph.” But, it doesn’t know what matters to me. It doesn’t lie awake turning over the same question for years. It doesn’t carry the history behind the word we when I talk about my family, or the quiet fear tucked under a sentence about illness.
That’s the part no model can do for me.
So, yes, I write with AI. Happily. Gratefully. It’s not a guilty secret; it’s just the current shape of my desk.
But, if any of this lands with you—if you recognize yourself in the tug-of-war between old myths of authorship and the reality of your actual life, that connection didn’t come from a model. It came from me choosing to say it out loud.
