How I Taught AI to Write in My Voice

Most AI output sounds like everyone else's AI output. Slightly formal. Generically structured. Technically correct but not something you'd actually say.

For a while, that was my experience too. Every piece I got back needed significant rewriting before it sounded like me. Then I started feeding the AI files about me.

Not blog posts or published content. Actual documentation about how I think, what I teach, what I never say, how I phrase transitions, what my frameworks are, what words are on my banned list. Seven markdown files covering my business context, my voice, my teaching style, my brand rules, and the patterns I rely on in every session.

I loaded those files into AI Studio in GHL and into my Claude projects. Now when I ask for content, I'm not asking a generic system. I'm working with a tool that's been briefed on who I am.

What changed

The difference wasn't subtle. Content that used to need heavy rewriting started coming out closer to what I'd actually say. Prompts that used to return corporate-sounding paragraphs started producing something I could work with in a first pass.

The reason is context. AI doesn't know who you are unless you tell it. A generic prompt gets a generic response. But a prompt submitted to a system that already knows your frameworks, your teaching philosophy, your signature phrases, and the words you'd never use — that's a different starting point entirely.

What I documented

The seven files I built cover these areas: my identity and background, my technical stack, how I like to work with AI, my core frameworks, my brand and design rules, my banned words and language constraints, and my voice and teaching style.

Some of those files took a few hours. Some took ten minutes. The voice file was the most involved because it captures specific patterns: how I open a story, what I say to move between ideas, how I structure a teaching sequence, what I avoid even when it sounds natural. The more specific you get, the more usable the output becomes.

How to build your own

Start with one file. Document three things: how you typically open a piece of content, five to seven phrases you'd never use, and the one teaching principle you come back to in nearly every session or post.

Load that into wherever you're doing your AI work. Ask it to write something you'd normally write and compare the output to what you'd say. Then add a second file covering your frameworks or your audience. Build gradually and test as you go.

This isn't about replacing your writing. You're always the author. It's about not having to undo AI's default assumptions every time you sit down to work. When the system is briefed correctly, the gap between what it drafts and what you'd actually say gets smaller. And smaller gaps mean faster work.

What's one thing you'd want AI to know about you before it writes anything with your name on it?