How to Use AI to Write Content That Actually Sounds Like You

I can tell within 3 sentences whether a LinkedIn post was written by AI. The giveaways: too many adjectives, no specific numbers, generic observations, and a conclusion that says nothing. The person didn't train the AI on their voice. They just asked it to "write a post about AI."

There's a better way. It takes 2 hours to set up and produces content that your audience will not be able to distinguish from your own writing.

Step 1: Gather Your Best Existing Content

Find 8-10 pieces you've written that got the best response: highest engagement posts, emails people replied to, blog posts that rank. These are the pieces where your real voice came through. Save them in a single document.

If you don't have 8-10 existing pieces, write 3-5 from scratch by hand without any AI assistance. The goal is capturing your natural expression. Raw, unpolished writing often captures voice better than your most edited work.

Step 2: Extract Your Voice Patterns With Claude

Paste all your content into Claude and use this prompt: "Analyze this writing and extract: 1) vocabulary patterns (words I use often, words I avoid), 2) sentence structure (length, complexity, rhythm), 3) structural patterns (how I open, transition, close), 4) tone characteristics (formal/casual, confident/hedging), 5) what topics and angles I gravitate toward. Format as a voice profile I can reference."

The output will surprise you. Claude will identify patterns you didn't know you had. Save this as your voice profile document.

Step 3: Build Your Production Prompt Template

Your template has three sections: the voice profile (paste the output from Step 2), the content brief (what you want to write, key points, target audience, platform), and the instruction ("Write in the voice described above. Apply all vocabulary, structure, and tone patterns precisely. Do not add em dashes, don't use the word 'revolutionary', and always include at least one specific number").

The specific prohibitions matter as much as the permissions. AI defaults to certain patterns (em dashes, certain adjectives, generic conclusions). You have to explicitly ban your pet hates.

What to Always Add Manually

Two things always require your hand: personal stories and specific numbers. AI can approximate your tone, but it doesn't have your memories. "A client told me last Tuesday..." or "we tested this on 847 contacts last month..." - those details have to come from you.

My workflow: AI writes the structure and prose, I add the specific stories and numbers, then I do a final pass to correct anything that still sounds generic. Total time: 15-20 minutes for a 500-word LinkedIn post that used to take 90 minutes.

The Pre-Publish Review Checklist

Before publishing anything AI-assisted: does it contain at least one specific number or date? Does it contain at least one personal reference (I, we, our)? Does the opening hook feel like something you'd actually say out loud? Does the conclusion drive toward something specific, not generic advice? If you fail any of these, fix before publishing.

We use this exact system for the AI Avengers content operation. Andy manages the production workflow, I provide the raw material (stories, numbers, angles), and AI handles the scaling. The full content system is something we teach inside AI Avengers Lab at aiavengers.team/lab.