Build an AI Content Engine That Sounds Like You

AI-written content is easy to spot: vague, over-polished, full of the same phrases. The fix is not a better “write me a post” prompt. The fix is a system where you supply the substance — opinions, experiences, examples — and AI does the heavy lifting on research, structure and repurposing.

The weekly loop

Block two sessions a week: a 30-minute input session and a 90-minute production session.

1. 1

   Pick three content pillars. Topics you can talk about from experience and that your audience cares about. Everything you publish should fit one of them.

2. 2

   Capture raw input all week. Voice memos, notes from calls, screenshots, things you changed your mind about. Transcribe voice notes (most phones and AI apps can do this) and drop everything into one doc.

3. 3

   Build a voice profile once. Give your assistant 5–10 pieces you wrote or said naturally (transcripts work well) and save the result. Paste it at the top of every writing prompt.

   Here are samples of how I write and speak: """ [paste 5–10 samples] """ Describe my voice so another writer could imitate it: sentence length, vocabulary, humor, how I open and close, words and phrases I use often, words I never use, how I explain technical ideas. Then list 10 "do not" rules to avoid sounding generic (e.g. banned phrases). Output as a reusable style guide.

4. 4

   Check what is rising in your niche. Spend ten minutes on trend research: what questions, tools or events are gaining attention this week? Match them against your pillars and your raw input — the overlap is your topic list.

5. 5

   Script from your input, not from scratch. Give the model your voice profile, the raw note and the angle. Ask for structure and hooks; keep your own examples and opinions intact.

Let the tools do the heavy lifting

Trend Radar covers step 4 — surfacing what is gaining attention in your niche. The Script Generator covers step 5 — turning your raw notes into hooks and scripts in your voice.

Want sharper research behind each piece? Pair this with the research prompt sequence: