You’ve asked ChatGPT to draft a blog post, an email, or a proposal. It came back polished, grammatically perfect, and completely soulless. Sound familiar? Here’s the thing: you can train ChatGPT to write like you, and it takes about ten minutes with a feature most people never touch.
The trick is custom instructions. Skip them and you’re stuck with ChatGPT’s default voice, which sounds like a polite robot that read too many corporate blogs. Set them up properly and the difference is honestly stunning.
Why ChatGPT sounds nothing like you
Default ChatGPT has a personality problem. It defaults to formal, balanced, slightly generic writing. That’s by design: OpenAI tuned it to be helpful and inoffensive to everyone. The side effect is that everything it writes, for everyone, comes out with the same flavor.
Meanwhile you have a voice. You might be funny, blunt, or fond of short punchy sentences. ChatGPT doesn’t know any of that unless you tell it.
The good news? Custom instructions exist exactly for this. You paste in a description of how you write, and ChatGPT follows it on every single reply. No remembering to re-prompt. No “hey, can you make it sound more like me?” at the end of every request.
I’ve been through this with my own content. Once you train ChatGPT to write like you, the drafts come back in your voice on the first try, and you’re editing instead of rewriting. That’s the entire promise, and it holds up when you set it up properly.
The 4 elements of your writing style
Before you can teach anything, you need to know what you’re teaching. Writing style breaks down into four parts, and this framework comes from a great Zapier piece by Matt Giaro that walks through the whole process. The method works because you train ChatGPT to write like you using your own words, not a generic prompt you found online.
Voice is how you sound on the page. Colloquial and chatty? Formal and precise? A decent custom instruction looks like: “Use this voice: Conversational and direct. Sounds like a smart friend explaining something, not a textbook.”
Tone is the attitude you carry toward the reader and the topic. Sarcastic, warm, no-nonsense, playful. Example: “Use this tone: Warm but no-nonsense. Encouraging without being cheerleader-y.”
Style covers your sentence-level choices. Minimalist? Long flowing sentences? Active voice always? Example: “Use this style: Short sentences. No jargon. One idea per paragraph. Active voice always.”
Structure is how you organize a piece. Linear storytelling? Open with tension, close with a takeaway? Example: “Use this structure: Opens with a tension or question, works through it step by step, closes with a single clear takeaway.”
Writing all four down gives ChatGPT a surprisingly complete picture. The clearer you get, the closer the output lands.
Step 1: Gather samples of your writing
Pull three to five pieces you’ve written that feel most like you. A couple of blog posts, an email you’re proud of, a social media thread. Text copies work best because ChatGPT can read them directly.
Don’t overthink the selection. You’re not looking for your best writing ever. You’re looking for writing that sounds like you talking.
Step 2: Let ChatGPT describe your style
This is the fun part, because you don’t have to describe yourself at all. Your job is to make ChatGPT do it.
Paste your samples into a fresh ChatGPT conversation and use a prompt like this:
“I’m going to paste examples of my writing. Analyze them and describe my writing style in detail: voice, tone, style, and structure. Be specific, and use examples from my text.”
Then paste your samples. What comes back is a style profile, usually a short paragraph per element, ready to use.
Read it before you save it. Sometimes ChatGPT flatters you. If it says your style is “elegant and sophisticated” and you write like a sports coach, correct it. You want the honest version, not the compliment.
Step 3: Paste the style guide into custom instructions
Open ChatGPT, click your profile, then Custom Instructions (it lives under Personalization in newer versions). Two boxes pop up: one about you, one about how you want ChatGPT to respond. The second box is where your style profile goes.
Paste the whole style description in there. That’s the entire setup. From now on, every reply, in every chat, follows your style. This is the step where you train ChatGPT to write like you for good, so don’t rush it. Reread the profile once before saving.
If you use custom GPTs, you can do the same thing: paste the profile into the GPT’s instructions. That’s useful when you want one voice for newsletters and another for client work.
Mistakes that ruin the setup
Most people fail at this in one of four predictable ways, and I’ve made most of them myself.
Being too vague. “Write like me” tells ChatGPT nothing. “Use short sentences, avoid jargon, and start paragraphs with a question when it fits” tells it plenty. The more specific your descriptors, the better the output. Vague in, vague out.
Describing who you want to be, not who you are. If you paste samples of your casual emails but describe your style as “formal and authoritative,” you’ll get formal writing that doesn’t match anything you actually wrote. Describe what’s on the page, not the writer you wish you were.
Copying someone else’s voice. Using a famous writer’s style profile can be fun for a one-off experiment. Using it for your real content is a mistake. Readers can smell borrowed voices, and consistency builds faster when the voice is actually yours.
Skipping the refinement pass. The first test draft is never perfect. People run it once, shrug, and never touch custom instructions again. Two rounds of feedback (“shorter paragraphs,” “fewer exclamation marks”) turn a close match into a real one. It’s five extra minutes.
How to keep it from drifting
Custom instructions aren’t static. Your writing changes over time, and the profile should too. Every couple of months, re-run the analysis with newer samples and refresh the instructions. It takes less than ten minutes and keeps the match tight.
Also worth knowing: custom instructions apply to every conversation by default. If you switch contexts a lot (personal chats in the morning, work emails in the afternoon), you might want a custom GPT with the work voice instead, so your personal replies don’t suddenly sound like your newsletter. That separation is why the two approaches complement each other.
Step 4: Test and refine
Give it a real task. Ask ChatGPT to rewrite one of your samples, or draft something new in your style. Compare it to the original. Close? Good. Off? That’s fine, this is a two-minute fix.
Tell ChatGPT what to adjust: more contractions, shorter sentences, less corporate phrasing. The profile updates, and the next output follows.
Here’s the honest part: ChatGPT gets close, not perfect. It still needs you to catch the parts only you’d catch, like inside jokes or your favorite weird phrasing. But you’ll spend minutes editing instead of hours starting from a blank page. That’s the whole point.
The takeaway
You can train ChatGPT to write like you in one sitting: gather samples, have it describe your style, paste the profile into custom instructions, then refine. Ten minutes of setup, and every draft from here on sounds like you instead of the robot.
Want to go further? Once your voice is locked in, you can plug it into a content system that never runs out of ideas, and keep an eye on the phrases that make writing sound like a bot so you don’t accidentally sound like one. Your voice is the one thing AI can’t buy. Might as well teach it to copy that.
