Your AI isn’t generic. Your inputs are.
Read time: 3 minutes
Your AI drafts read fine. They also read like everyone else’s AI drafts. Flat rhythm, safe words, zero fingerprint.
You’ve rewritten the prompt five times this month. The voice still isn’t there.
You’ve tried longer prompts, tone instructions, “write like Naval” requests. The model complies for a paragraph, then drifts back to generic.
Even pasting old posts barely helps. The model skims tone and misses substance.
Instructions describe a voice. Material carries it.
A prompt tells the model what to sound like. It has no sample of your voice to pull from.
Here is the system I run every week.
The voice archive
Collect your raw material
Feed the model real language before you ask it to imitate you. Pull from the places you already sound like yourself.
- Voice memos where you explain an idea out loud
- Slack replies and DMs, unedited
- Sales call transcripts, the off-script moments
- Old posts and emails that got engagement
Drop it all into one Notion database. Nothing polished. Raw signal only.
↳ Add to it whenever you catch yourself sounding like you.
Write your voice file
Raw material alone doesn’t transfer. Turn those samples into an explicit rulebook: a sentence length cap, a banned words list, and three real paragraphs pulled straight from the archive.
Use this prompt to speed up the extraction:
> Read these 10 raw samples. Output a rules list: average sentence length, 15 overused words, 15 I’d never say. Add 3 sentences that sound most like me.
↳ Mine holds banned words, sentence caps, and sample paragraphs.
Load it into every draft
“Write like me” gives the model nothing. It has no memory of your voice between chats.
So load the voice every time. Paste the **voice file** first, then add a **source brief** with the facts for this draft.
This exact setup runs every issue of this newsletter. Voice file, banned-phrase list, real source notes. Stored in Notion, loaded fresh into Claude on every draft.
↳ Result: drafts that don’t need a rewrite pass for voice.
The payoff
Build the voice file once. Every draft after sounds like you.
No rewrite pass to sound like yourself again. Paste, generate, ship.
🧭 You can install your own voice. The system that turns it into revenue is where operators stall.
Eric Melillo
Find me on X, LinkedIn, Instagram,
or Book 1:1 Strategy Call
PS – Whenever you’re ready, here are 3 ways I can help you:
- Free Resources – Get swipe files, templates & AI prompts to accelerate your growth.
- Online Courses – Learn step-by-step from proven frameworks to build and scale your creator business.
- Book 1:1 Strategy Call – Being stuck is a choice. In 30 minutes, we’ll map your next 3 moves to get you unstuck. No cost, no strings. Creators who win don’t wait.