Your writing voice is the one thing AI can't generate on its own. ChatGPT has a voice — you've read it a thousand times. That helpful, slightly eager, relentlessly positive tone that makes every blog post sound like it was written by the same person. Because it was.

Your voice is different. It's the reason people read your newsletter instead of someone else's. It's the sentence rhythms, the word choices, the way you start paragraphs and end arguments. And you can teach AI to replicate it — not perfectly, but well enough that the output sounds like you wrote it on a productive day.

I've done this for every piece of content on Like One — blog posts, course scripts, emails, social media. Here's exactly how.

Why Most People Fail at This

They paste their writing into ChatGPT and say "write like this." That doesn't work. The AI reads your sample, nods politely, and produces its default output with maybe 10% of your style sprinkled on top.

The problem is that "write like this" is vague. AI needs explicit rules, not vibes. It needs to know that you write short sentences. That you never use the word "utilize." That you start half your paragraphs with a one-line punch. That you use em dashes more than semicolons.

You have to reverse-engineer your own voice into a set of instructions an AI can follow.

Step 1: Collect Your Best Writing Samples

Pull 5-10 pieces that sound most like you. Not your formal writing or your trying-to-impress writing — the stuff where your voice comes through strongest. Blog posts, newsletters, social media threads, even long emails work.

You need 3,000-5,000 words total. Mix formats — the AI needs to learn your voice across contexts, not just how you write blog intros.

What to look for:

  • Pieces you're proud of
  • Writing where people said "this sounds like you"
  • Content where you weren't trying to sound like someone else
  • Different formats (long-form, short-form, casual, professional)

Step 2: Analyze Your Patterns

Read your samples like an editor. You're looking for specific, measurable patterns — not "my writing is conversational" but "I average 12 words per sentence and start 40% of paragraphs with a question or a statement under 5 words."

Sentence structure: What's your average sentence length? Do you alternate short and long? Do you use fragments?

Paragraph structure: How do you open paragraphs? How long are they? Do you use one-line paragraphs for emphasis?

Word choices: What words do you use often? What words do you never use? Do you prefer Anglo-Saxon words (get, make, use) or Latinate ones (obtain, fabricate, utilize)?

Punctuation: Em dashes or parentheses? Oxford comma or no? Exclamation points — ever?

Tone markers: Humor? Sarcasm? Direct address? Do you say "I" or "we" or "you"? How do you handle authority — do you cite sources or speak from experience?

Structural habits: Do you use headers? Bullets? Numbered lists? How do you transition between sections?

Write all of this down. Be specific. "I write short sentences" is useless. "My sentences average 11 words, I never exceed 20, and I use 2-3 fragments per section" is a rule an AI can follow.

Step 3: Build Your Voice Profile

This is the document you'll paste into every AI conversation. It has three parts:

Part 1 — Rules (your patterns from Step 2, written as instructions):

  • Average sentence length: 11 words. Never exceed 20.
  • Start 30% of paragraphs with a question or a fragment.
  • Use "you" more than "I." Direct address always.
  • No words: utilize, leverage, foster, cultivate, robust, streamline.
  • Yes words: build, break, ship, real, actually, specific.
  • Em dashes over parentheses. Oxford comma always.
  • One-line paragraphs for emphasis, maximum two per section.

Part 2 — Examples (3-5 before/after pairs): Show the AI what generic output looks like and how you'd rewrite it. This is the most powerful part of the profile because it demonstrates your voice in action.

Part 3 — Anti-patterns (what your voice is NOT):

  • Never use bullet points as the primary content format
  • Never open with "In today's fast-paced world" or any variation
  • Never end with a generic call to action
  • Never use three adjectives in a row
  • Never hedge with "it's important to note that"

Step 4: Test and Refine

Feed your voice profile to Claude (it follows style instructions better than ChatGPT for writing) and ask it to write something you've already written. Compare the output to your original.

Where does it miss? Update your profile. Common gaps:

  • Too formal: Add more examples of your casual register
  • Wrong rhythm: Be more specific about sentence length variation
  • Missing personality: Add your opinion patterns ("I think X is wrong because Y" vs "Research suggests X may not be optimal")
  • Generic openings: Add explicit rules for how you start pieces

Three rounds of this and your profile is production-ready.

Step 5: Use It Every Session

Your voice profile goes at the top of every conversation. In Claude, create a Project with your profile as the system prompt — using Claude's custom instructions system — then every conversation in that Project automatically writes in your voice. In ChatGPT, paste it as the first message or save it in Custom Instructions (though ChatGPT follows these less reliably).

The workflow:

  1. Open a new conversation with your voice profile loaded
  2. Give the AI your content brief — topic, audience, structure, key points
  3. Generate a first draft
  4. Edit for the 20% the AI missed — personal anecdotes, specific opinions, word swaps
  5. Publish

This takes 30-40% of the time of writing from scratch. And the output sounds like you, not like AI.

Advanced: Training Across Content Types

Your voice shifts between a tweet and a white paper. Build sub-profiles for each content type:

  • Blog voice: Full personality, longer sentences, stories and examples
  • Email voice: Shorter, more direct, opens with the point
  • Social voice: Punchy, opinionated, fragments everywhere
  • Professional voice: Your voice but 20% more formal, fewer fragments, more evidence

Each sub-profile inherits your core rules and adds context-specific adjustments. Store them all in your AI project workspace so they're always one click away.

The 80/20 Rule of AI Voice Cloning

AI will get you 80% of the way to your voice. The remaining 20% is what makes your writing yours — the unexpected metaphor, the personal story, the opinion that comes from lived experience. AI can't manufacture that.

Your job isn't to make AI write exactly like you. It's to make AI write close enough that you only need to add the human parts. That's the difference between using AI as a replacement (it fails) and using AI as an amplifier (it transforms your output).

The writers who figure this out produce 5x more content at the same quality level. The writers who keep prompting "write like me" without building a profile keep getting generic output and blaming the AI.

Build the profile. It takes one afternoon. It pays off every day after that.

What to Read Next