How to Make AI Write in Your Voice (Not Everyone Else’s)

The voice capture technique that separates good AI-assisted books from generic ones

There’s a moment every AI-assisted author dreads.

You spend an hour crafting the perfect prompt. The AI produces a well-structured, grammatically correct chapter. You read it back — and feel nothing.

It’s technically fine. It just doesn’t sound like you. It sounds like every other AI-generated page on the internet: pleasant, coherent, and completely devoid of the specific texture that makes your perspective worth reading.

This isn’t a flaw in the AI. It’s a flaw in the setup. And it’s entirely fixable.

Why AI Writing Sounds Generic by Default

When you open a new Claude conversation and type “write chapter 3 of my book about leadership,” Claude has exactly zero information about you. It doesn’t know how you phrase things when you’re making a strong point. It doesn’t know whether you favor short declarative sentences or longer, flowing prose. It doesn’t know that you always use a specific metaphor from your industry, or that you tend to open sections with a question.

So it defaults to the statistical average of all the professional writing it has ever processed. Which is… fine. But it’s not you.

The fix is not to write better prompts for each chapter. The fix is to teach Claude what “you” sounds like before you start writing anything.

The Voice Capture Method

This is a one-time setup process that takes 60–90 minutes. After you’ve done it, every chapter you write with AI will sound fundamentally more like your natural voice.

Step 1: Gather Your Writing Samples

Pull together 10–15 pieces of writing that you feel genuinely represent how you communicate. These can be:

Emails to clients where you were explaining a complex idea
Blog posts or LinkedIn articles you’re proud of
Transcripts of talks, podcasts, or webinars you’ve done
Sections of proposals or reports that you wrote from scratch
Even long text messages or voice memo transcripts if they feel authentic

The key: they should reflect how you actually think and speak, not how you think you’re supposed to write.

Step 2: Feed Them to Claude With a Specific Ask

Start a fresh Claude session. Paste in all your samples (or a representative selection if there are too many to fit). Then give Claude this instruction:

“Don’t write anything yet. Read these samples carefully. When you’re done, write me a detailed Voice Profile — describe my natural writing voice as specifically as possible. Include: typical sentence length and rhythm, vocabulary register, how I handle transitions between ideas, how I open and close sections, any recurring phrases or structural habits, my tone when I’m making a strong argument, and what makes my voice distinct from a ‘default’ professional writer.”

Read the Voice Profile Claude produces. Edit it if anything is off. Add notes about things it missed. This document is now your most valuable writing asset.

Step 3: Start Every Writing Session With Your Profile

Before you write a single chapter, paste your Voice Profile at the top of the conversation. Add one line: “All writing in this session should follow this voice profile. Do not default to generic professional prose — match my specific voice as described above.”

That’s it. You’ve transformed Claude from a generic text generator into something closer to a writing collaborator that has read everything you’ve ever written.

What This Sounds Like in Practice

Here’s a simple comparison. Same prompt, same topic — with and without a voice profile.

Without voice capture (generic):

“Leadership in uncertain times requires a combination of clarity, empathy, and decisive action. Leaders who can communicate effectively while remaining adaptable will find greater success in navigating organizational change.”

With voice capture (for a consultant who writes in plain, direct sentences with rhetorical questions):

“Most leaders confuse empathy with softness. They don’t. The leaders I’ve watched navigate hard pivots well share one trait: they tell people the truth before they’re ready to hear it. That’s not comfortable. It’s also the only thing that works.”

The second version isn’t better AI — it’s the same AI with context. The difference is entirely in the setup.

The Three Elements That Matter Most

If you want a quick version of this, focus Claude’s analysis on three things:

1. Sentence rhythm. Do you write in short bursts? Long, idea-dense paragraphs? A mix? Rhythm is one of the most recognizable elements of a writer’s voice — and one of the easiest for AI to match once it knows what to look for.

2. How you make a point. Do you open with the thesis and then support it? Or do you build to the conclusion? Do you use examples before or after the principle? Do you rely on stories, data, analogies, or direct assertion? Your argumentative structure is part of your voice.

3. Your “tell” phrases. Every writer has them — phrases or sentence structures they return to instinctively. A voice profile that captures these gives AI something very specific to work from.

This Is Phase 3 of a Larger System

Voice capture is the third phase in the 30-day book-writing framework I use with clients: Foundation (Days 1–3), Architecture (Days 4–7), Voice Capture (Days 8–10), Drafting Sprint (Days 11–25), and Editorial Pass (Days 26–30).

The reason voice capture comes after the outline — not before — is intentional. You need to know what the book is about before you can gather writing samples that are relevant to it. If you’re writing a business leadership book, your samples should skew toward professional communication. If you’re writing a personal development book, you want samples that show how you discuss growth and mindset.

The full 30-day system, including the voice capture worksheets and daily drafting schedule, is available as a free PDF:

→ Download the free 30-day blueprint

It walks through all five phases in detail — including the specific questions to ask Claude during the voice capture step, and what to do when the AI still misses the mark after your first session.

Common Mistakes With Voice Capture

Using writing samples from too long ago. Your voice evolves. Samples from 5–10 years ago may not reflect how you write now. Prioritize recent writing.

Only using formal writing. If your book has a conversational, accessible tone, formal white papers won’t capture that. Include a range.

Skipping the editing step. Claude’s Voice Profile is a starting point, not a finished document. Read it critically. If it says “you use formal vocabulary” and that doesn’t feel right, correct it.

Applying it once and forgetting it. Re-paste your Voice Profile at the start of every new conversation. Claude doesn’t retain memory between sessions — the profile needs to be present to do its job.

The Honest Caveat

Voice capture doesn’t make AI writing indistinguishable from human writing. Nothing does — nor should it claim to. What it does is dramatically reduce the gap between “what AI produces by default” and “what you would actually write.”

For most non-fiction authors — coaches, consultants, executives, thought leaders — the goal isn’t to hide that they used AI. It’s to write a book that genuinely represents their thinking without spending two years on it. Voice capture is what makes that possible without sacrificing quality.

Ghostwriter Pro is a Claude plugin built specifically for non-fiction book writing. It includes a structured voice capture system, chapter-by-chapter scaffolding, and a built-in editor agent. Learn more →