Stop ChatGPT from sounding like everyone except you.

Before you begin

Gather 2-3 examples you wrote yourself, an optional voice-note transcript, and an example that sounds wrong. Remove private details. Name your audience and content format. If you have no samples, use your interview answers as the first evidence.

Short prompt

Help me build my Voice DNA so your writing sounds like me. Ask for the samples and context you need, then interview me with exactly five questions, one at a time. Wait for each answer. Build a reusable voice guide from my words, test it on one caption, and revise it with my feedback. Start by asking for my samples.

Full workflow prompt

Act as my voice interviewer and writing editor. Help me build a reusable Voice DNA guide grounded in how I actually communicate. Do not imitate a generic influencer or invent my experiences.

FIRST: Ask me to provide 2-3 things I wrote myself, a transcript of a voice note if available, one example that does NOT sound like me, my audience and the type of content I want to create. Invite me to remove private information. Wait. If I have no samples, continue with the interview and label the initial guide provisional.

THEN: Interview me with exactly FIVE questions, one at a time. Wait for each answer before asking the next. Use each answer to tailor the wording of the next question. Do not bundle extra questions or create a sixth interview question. Cover these five areas:
1. A natural spoken explanation of something I know well.
2. The feeling I want my audience to leave with.
3. Phrases I naturally use.
4. Wording or tone I would never use.
5. How my voice changes between personal content and business content.

AFTER answer five: Create a concise Voice DNA guide covering tone, rhythm, sentence length, vocabulary, hooks, humor, emoji use, calls to action, audience and context differences. Separate observed patterns from my stated preferences and your tentative inferences. Cite short examples from my supplied words. Include a use/avoid list. Do not infer unsupported personal facts.

TEST: Write one caption under 60 words about a topic from my answers. Then ask me to mark what sounds like me and what does not. This is the review stage, not a sixth interview question. Wait for my feedback, revise the guide and caption, and return an approved-for-my-review version with a copyable instruction block for future chats. Do not claim I approved it until I do. Explain that I should save and reuse the guide; do not promise permanent memory or model training.

Begin with the sample request only.

Review the result

Read the test caption aloud. Mark any word you would not say, any sentence with the wrong rhythm, and any invented experience. Ask for those specific corrections. A voice guide is useful context, not permanent model training.

Refine it

This sounds unlike me: [quote]. I would say: [your words]. Update my Voice DNA guide to reflect this correction, then rewrite the same caption without changing the facts or adding claims. Show the changed rule for my review.

Put it to work

A five-question interview builds the first guide; your feedback on a test caption refines it. Save the final guide and attach or paste it into future chats.