You Can Use AI Without Losing Your Voice (Here’s How)

Six months ago, I asked ChatGPT to “improve” a paragraph I’d written. The result looked professional, technically correct, even elegant in places. But when I read it, something felt wrong. My sentence about a stomach full of angry bees had become “significant anxiety and nervous anticipation.” The punchy two-beat rhythm I’d written turned into a long academic crawl.

Every edge that made the paragraph mine had been filed down.

AI doesn’t erase your voice on purpose. But left unchecked, it will sand down everything distinctive about your writing until you’re left with something smooth, confident, and utterly forgettable.

You’ve probably felt this if you’ve used ChatGPT for anything creative. You ask it to brainstorm or rewrite a section that’s not working. The output looks fine—good, even. But when you come back to it later, you realize it doesn’t sound like you anymore. It sounds like everyone else who’s using the same tool the same way.

This isn’t your fault. It’s not even AI’s fault, really. It’s a predictable outcome of how these models work—and once you understand the mechanism, you can protect against it.

The AI models we use for writing are prediction engines. They’ve been trained on billions of words from books, articles, and websites, learning patterns—what word typically follows another, what phrases show up together, what structures recur. Then they predict the statistically most likely next word based on all that data.

This makes them powerful for pattern matching and summarizing. But there’s a side effect: because these models learn from the average of everything, they produce the average. They default to a smooth, confident, utterly generic “announcer” voice. All the unique, jagged edges—your quirks, your rhythm, the worldview that shapes how you see things—get sanded down.

I call this the flattening effect.

Here’s an example. I wrote this sentence:

“Expertise is raw material. A book is a constructed object. Most writers fail because they mistake the pile of bricks for the house.”

When I asked ChatGPT to “rewrite this to be clearer and more professional,” it gave me:

“It is important for writers to understand that expertise alone is insufficient for creating a successful book. The process of writing a book is akin to building a structure, where the author’s knowledge serves as the foundational material.”

Notice something?

The second version isn’t wrong. The information is there. But it has no energy, no conviction. The punchy rhythm is gone. The specific image—”pile of bricks”—becomes vague “foundational material.”

That’s the flattening effect in action.

A few months ago, I asked ChatGPT to write a fantasy novel in the style of Tolkien’s The Fellowship of the Ring. I gave it a detailed prompt—humble hero, dangerous object, themes of sacrifice and loss. It delivered 200 pages in under an hour.

The prose was technically perfect. Poetic language, vivid imagery, all the Tolkienesque elements you’d expect. It could generate “a comet blazing gold and furious crimson” tearing across the sky with crystalline accuracy. But it couldn’t tell you why that moment should make a reader’s breath catch. It arranged words beautifully while saying nothing that matters.

Patterns aren’t the same as meaning. Style isn’t substance.

This is the central danger for creative writers. AI defaults to what’s statistically probable—which is, by definition, what’s already been done. It gives you the smooth center of the distribution, not the jagged edges where your unique voice actually lives.

Most writers worry about AI replacing them. That’s not the real risk.

The real risk is AI slowly eroding what makes your writing yours—until you open your manuscript six months later and don’t recognize your own voice.

What I wish I’d known earlier: you don’t need to avoid AI. You need a system that lets you use AI’s speed and power without sacrificing what makes your writing uniquely yours.

The core principle is straightforward: Your process, your judgment, your voice.

AI can be a powerful assistant within your established creative process. But it can’t be allowed to become the process itself.

When I started writing The Maker’s Heir—a novel about a robot with Alzheimer’s—I didn’t open ChatGPT first. I walked around a pond near my house in March, thinking about my father and brother, both dead from cancer. I thought about what makes a life matter when memory starts to fail.

That question—the one that eventually became a story about memory, grief, and what it means to be human—didn’t come from an AI prompt. It came from loss and walking and thinking. ChatGPT could have written cleaner sentences than I did in my first draft. It probably could have described the setting more eloquently. But it couldn’t have found the reason to write the story.

AI generates content. You generate meaning. That’s the distinction that matters.

After a year of experimenting—making mistakes, throwing out 20,000 words that didn’t work, figuring out what actually does—I’ve distilled the craft-first approach into three core protections:

You can’t protect what you can’t define. Most writers struggle to articulate what makes their voice unique. Ask them to describe their style and you’ll get vague answers: “conversational,” “clear,” “engaging.” Those descriptors aren’t specific enough to protect against AI’s flattening effect.

The Voice Anchor framework has three components:

• Core Essence: Your fundamental posture on the page—not your credentials, but who you ARE when you write

• Linguistic Fingerprint: How you build sentences (rhythm, structure, patterns)

• Voice Boundaries: What you NEVER sound like (forbidden registers and constructions)

Once you can articulate these explicitly, you can evaluate any AI output against your actual voice. When something doesn’t match, you’ll recognize it immediately.

Most writers ask AI to write for them. That’s the wrong approach.

The I-R-C framework teaches you to use AI as a thinking partner instead:

• Ideation prompts: Brainstorm options—AI generates possibilities, you decide which to pursue

• Research prompts: Get explanations customized to your specific needs, not generic Wikipedia answers

• Challenge prompts: Red-team your work—AI diagnoses problems, you decide how to fix them

The pattern stays consistent: AI expands possibilities and stress-tests logic. You make every creative decision.

I always write the first draft. No exceptions.

This is the line you never cross. It’s non-negotiable because this is where your voice lives and where you make the connections only you can make.

When I drafted The Maker’s Heir, I didn’t consult AI. I wrote alone. The prose was rough—clunky transitions, repeated words, sentences I knew I’d have to fix later. But the voice was mine. The connection between my father’s death and the AI character’s memory loss came from me, not from a prompt. The emotional truth of the ending came from walking and thinking, not from ChatGPT.

AI could have written cleaner sentences in that first draft. But it couldn’t have found the reason to write the story. That’s what you protect with the Golden Rule.

The complete system follows a 4-stage workflow:

Stage 1: Idea & Positioning → You + AI Ideation (expand possibilities)

Stage 2: Outlining → You + AI Challenge (stress-test structure)

Stage 3: Drafting → YOU ALONE (The Golden Rule—no exceptions)

Stage 4: Revising → You + AI Challenge (diagnose problems, you fix)

Notice Stage 3—that’s where AI goes away completely. This isn’t arbitrary. Each stage has a different relationship with AI, depending on the kind of thinking you’re doing. In Stages 1 and 2, you’re exploring and structuring, so AI can help expand options. In Stage 3, you’re creating—this is where voice lives, so AI stays out. In Stage 4, you’re evaluating, so AI can help diagnose problems with what you’ve already created.

You can use AI to amplify your creativity without flattening your voice. But you need three things:

1. Voice Anchor to define what you’re protecting

2. I-R-C prompts that preserve creative authority

3. Golden Rule that keeps AI away from drafting

I’m building a workshop that teaches all three—with hands-on exercises, fillable templates, and real examples from my own writing process. It’s called The Craft-First AI Workflow, and it covers everything you need in five modules (about 3 hours total).

After the workshop, you’ll have a complete Voice Anchor you can use to evaluate all AI output, customizable I-R-C prompt templates, a 4-stage workflow showing exactly when to use AI and when to work alone, and a one-page Personal AI Protocol that serves as your governance system for all AI decisions.

The goal: Use AI as a thinking partner without losing what makes you YOU.

Subscribe to get notified when it’s available. I’ll also be making early modules available to paid subscribers.

If you’re on the fence, try this exercise first:

Take a sentence from your own writing—something that sounds 100% like you. Paste it into ChatGPT with this prompt: “Rewrite this to be clearer and more professional.” Look at what comes back and notice what gets lost. That’s what you’re protecting against.

Once you see the flattening effect in your own work, you can’t unsee it. You’ll understand why having an explicit defense mechanism matters.

Speed without direction just gets you lost faster. You don’t need to write faster with AI. You need to write work that matters—work that connects with readers because it says something only you can say. That’s the work this workshop protects.