How to make ChatGPT
sound like you.
You’ve probably tried. Custom instructions, “write this in my style,” pasting an old email into the prompt. The output gets closer — less corporate, fewer bullet points — but a colleague who knows your writing can still tell. Here’s why, and what actually closes the gap.
What people try first
Custom instructions
“Write casually, avoid em-dashes, keep sentences short.” Helps with tone, but rules describe a category of writer — not you. Two people with the same instructions get the same voice.
Pasting samples in the prompt
Better — the model imitates surface features for a few paragraphs, then drifts back to its defaults. And you never know how close it actually got.
Editing every output by hand
Works, and that's the problem: ten minutes per email quietly cancels most of what the AI saved you.
Why prompting alone falls short
A writing voice isn’t a set of rules you can list — it’s a bundle of measurable habits: how long your sentences run and how much they vary, which greetings and sign-offs you reach for, whether you write “thanks!” or “thank you,” how often you use commas, parentheses, ellipses. You can’t describe all of that in a prompt because most of it is invisible to you — it’s just how you type.
A language model asked to “write like me” guesses at these habits from a small sample, holds the imitation for a few paragraphs, and then reverts to the patterns it was trained on. Without measurement, neither you nor the model knows how far off it landed.
What measuring your style does differently
Shlokah computes your style from writing you already have — pasted emails, WhatsApp exports, notes — into concrete numbers and patterns. Every generation is then scored against those numbers, and when a draft measurably misses (too formal, sentences too uniform, phrases you’d never use), it’s revised automatically before you see it.
The result isn’t “ChatGPT with a personality prompt.” It’s your rhythm and vocabulary held to a measured standard, with a style-match score on every output so you can trust it before you hit send.
Frequently asked questions
Can ChatGPT learn my writing style?
Partially. With custom instructions or pasted samples, ChatGPT will adjust tone and formality, but it drifts back to its defaults within a few paragraphs and can't verify how close it got. It approximates a vibe rather than measuring your actual rhythm, vocabulary, and punctuation habits.
What prompt makes ChatGPT write like me?
The best prompt-only approach: paste 2-3 samples of your writing and ask it to 'match the sentence length, punctuation, greetings, and vocabulary of these samples exactly — do not use any phrasing that doesn't appear natural to this writer.' Expect roughly 60% fidelity, degrading as the conversation continues.
How is Shlokah different from prompting ChatGPT with my writing?
Shlokah computes concrete stylometrics from your samples — average sentence length, punctuation frequencies, characteristic phrases, AI-tell patterns you never use — then generates, measures the result against those numbers, and revises automatically when it misses. You see a style-match score instead of guessing.
Do I have to connect my email or accounts?
No. You paste or upload only the writing you choose — past emails, WhatsApp exports, notes. Nothing is scraped, samples stay private to your account, and you can delete them anytime.
Stop prompting. Start measuring.
Paste a few things you’ve written and hear the difference — 3 free generations in your own voice, no card required.
Build your style profile