Prompting

How to Make ChatGPT Match Your Writing Style (Step-by-Step Guide)

You ask ChatGPT for a draft. It comes back clean, correct and instantly forgettable. Every paragraph is the same length. Every sentence is “balanced.” Somewhere in there is the word “delve.”

So you type: “Write it in my style.”

Nothing much changes, and that’s not ChatGPT’s fault. “My style” means nothing to a model that has never read your work. It needs evidence: real examples of how you write, clear rules about what you avoid, and a place to save those rules so they don’t disappear when you open a new chat.

This guide walks you through the whole process, from collecting samples to setting up ChatGPT’s built-in writing style feature. Every step comes with a prompt you can copy.

Quick answer: Give ChatGPT 3 to 5 samples of your typical writing. Ask it to analyze your patterns before it writes anything. Turn that analysis into concrete rules, save them in a Style File, Custom Instructions, a Project or the writing style setting, and test the result on a real task. Then keep refining.

What “Your Writing Style” Actually Means

Before you can teach ChatGPT your style, it helps to know what it’s made of. Most writing styles come down to four elements:

  • Voice: your personality on the page. Are you casual and playful, or calm and authoritative? Your voice stays the same from piece to piece.
  • Tone: the mood of a specific piece. Your voice might always be friendly, but your tone can be urgent in a product update and relaxed in a how-to.
  • Style: the small choices. Word choice, sentence length, rhythm, and whether you reach for analogies, humor or data.
  • Structure: how you organize ideas. Do you open with a story or a bold claim? Do you love lists? How do you end?

Keep these four in mind. You’ll use them to judge every analysis and every draft ChatGPT gives you.

Step 1: Collect the Right Writing Samples

This step decides everything else. Bad samples produce a confused style.

Choose 3 to 5 pieces (about 1,000 to 2,000 words total) that sound like you. Then check them against these rules:

  • Use typical writing, not your one-off hit. You want ChatGPT to learn how you usually write, not the one viral post you rewrote twelve times.
  • Match the format. Want blog posts? Use blog posts. Want emails? Use emails. A model trained on your LinkedIn posts will struggle to write your newsletter.
  • Use only your own writing. No AI-assisted drafts, no text a freelancer or editor wrote for you. Mixing authors gives you a blurry voice that belongs to nobody.
  • Prefer lightly edited text. Heavily edited copy often reflects your editor more than you.

Label every sample. Put your samples in one document and separate them clearly with headings like “Sample 1,” “Sample 2” and so on. ChatGPT needs to see where each text starts and ends, or it may blend them into one.

Going deeper? For a stronger result, use 5 to 10 samples per format you write in. Pick the pieces that best represent how you write, not just the ones that performed best.

Step 2: Ask ChatGPT to Analyze Your Style (Before It Writes)

The most common mistake is pasting samples and immediately asking for a draft. When you do that, the model starts performing a style before it has studied the evidence.

Ask for an analysis first:

Read the writing samples below. Do not write anything new yet.

Analyze my writing and identify recurring patterns in:
- Voice and tone
- Sentence length and rhythm
- Word choice: words and phrases I use often, and ones I seem to avoid
- How I open and how I close
- How I build an argument or explain an idea
- Use of examples, stories, humor or data
- Formatting habits (headings, lists, paragraph length, punctuation)

Give the results as bullet points.

Samples:
[PASTE YOUR SAMPLES]

Repeat this with each sample if you have time, then look for patterns. If ChatGPT describes you as “direct,” “practical” and “slightly playful” in three different analyses, that’s real. If a description appears once and doesn’t feel like you, ignore it.

You know your writing better than any model. Treat the analysis as a first draft of your identity, not the final word.

Step 3: Turn the Analysis Into Concrete Rules

An analysis is not useful until it becomes something you can reuse. That means rewriting it as clear, specific rules.

The key is to delete vague descriptions and keep concrete ones:

❌ Vague (delete)✅ Concrete (keep)
Uses a conversational toneTalks to the reader as “you” and asks one question per section, max
Writes clearlySentences are usually under 20 words; one idea per paragraph
Engaging introsOpens with a specific problem or a short scene, never a definition
Uses examplesGives a real-life example after every main point
Strong endingsEnds with one short, flat sentence that states the takeaway

You can ask ChatGPT to do the first pass for you:

Turn your analysis into a Style Guide of 10 to 15 rules.
Each rule must be specific enough that another writer could follow it
without seeing my samples. Remove vague words like “engaging”,
“clear” or “conversational” unless you explain exactly what they mean.

Read the result, fix anything that’s wrong, and save it in the same document as your samples. This Style Guide is the most valuable thing you’ll create in this process. We’ll turn that document into your reusable Style File in Step 5.

Step 4: Tell ChatGPT What to Avoid

Your style isn’t only what you do. It’s also what you’d never do. A “never” list is often what removes the robotic feel fastest.

Ban the classic AI tells

Start with the patterns readers have learned to spot:

Never use these words or phrases: delve, tapestry, landscape, realm,
robust, leverage, seamless, elevate, unlock, game-changer,
“in today’s fast-paced world”, “it’s important to note”,
“whether you’re a beginner or a pro”, “in conclusion”.

Avoid:
- Em dashes (—). Use a period to start a new sentence or a comma
  to continue it.
- Starting paragraphs with “Additionally,” “Furthermore” or “Moreover”
- Lists of exactly three adjectives
- Rhetorical questions in every section
- Summaries at the end that repeat the whole article

Find ChatGPT’s own favorite words

Here’s a trick most guides skip. After a few drafts in the same chat, ask ChatGPT to show you which words it keeps repeating:

Analyze only your own responses in this chat (not my messages).
List the 20 words you used most often, excluding common stop words.
Show a table with: rank, word, frequency, and one example sentence.
Then stop.

If you see words like “clarity,” “structure” or “ensure” near the top, and they’re not words you use, add them to your banned list.

Step 5: Save Your Style So You Don’t Repeat Yourself

A style guide that lives in one chat dies in that chat. Save it somewhere ChatGPT will see it every time. You have five options:

OptionBest forKeep in mind
A Style FileAny plan and any AI tool, including the free version, Claude and othersYou attach it manually at the start of each chat
Custom InstructionsOne main voice for almost everything you writeThe fields are short, so save a condensed version of your rules
ProjectsSeveral voices (your blog, a client, your newsletter)Add your samples and full style guide as files, and short rules as project instructions
Writing style settingLetting ChatGPT learn from your actual documentsNewer feature; availability and menus may vary
MemorySmall facts about you and your workLeast reliable for style, since you don’t fully control what it remembers

Your Style File (works anywhere)

The simplest option needs no settings at all. Create one document called something like “My Style File” with two parts:

  1. Your Style Guide from Step 3, at the top
  2. Your labeled samples (“Sample 1,” “Sample 2”…) underneath

When you start a new writing task, attach the file and use a prompt like this:

Attached is my Style File: a guide to my writing style plus samples
of my writing. Read it before you write. Then write [your task] so
it sounds like I wrote it. Follow my Style Guide, match the
structure and tone of my samples, and do not copy their phrases.

Keep it growing. Every time you write something that feels very you, add it to the file. Swap out older samples that no longer sound like you. The file gets better as your collection grows.

Because it’s just a document, your Style File works in any AI tool. It’s also the document you’ll connect to the built-in writing style setting described below.

Custom Instructions

Open ChatGPT’s settings, go to Personalization, and find Custom Instructions. You’ll typically see two areas: one for context about you and one for how ChatGPT should respond.

  • About you: your audience and work. For example: “I write practical blog posts for freelance content writers who use AI.”
  • How to respond: your condensed rules, using a simple format:
Voice: friendly, direct, slightly playful
Tone: encouraging but honest; no hype
Style: short sentences, plain English, one idea per paragraph,
a real example after every main point
Structure: open with a specific problem, use clear H2 sections,
end with one short takeaway sentence
Never: em dashes, “delve”, “landscape”, “game-changer”,
corporate jargon, motivational fluff

Tip: If you only use ChatGPT for writing, you can turn off tools you don’t need, like image generation or code, in the same settings. It keeps responses focused.

Projects

If you write in more than one voice, Projects are the better choice. Create one project per voice, such as “Blog,” “Newsletter” or “Client: Acme.” In each one:

  1. Upload your Style File (samples and full Style Guide) as project files.
  2. Put your condensed rules in the project instructions.
  3. Start every writing chat for that voice inside its project.

Now every new chat starts with your voice already loaded, and your client’s brand voice never leaks into your personal blog.

ChatGPT’s built-in writing style setting

ChatGPT now has a dedicated writing style option in its Personalization settings. Instead of describing your style, you point ChatGPT to your own documents, for example a file in a connected Google Drive, and it studies your sentence structure, word choice and formatting habits directly.

A few tips to get the most from it:

  • Use your Style File (or a dedicated document with only your own, typical writing). One user reported good results with a file of around 100 of their own posts.
  • Point ChatGPT to that file. Add an instruction like: “When I ask you to write in my style, use the document ‘My Style File’ as the main reference.” Otherwise it may pull from random notes and old drafts in your Drive.
  • Update the file over time. Your style changes, and an outdated sample set will hold you back.

Note: ChatGPT’s settings and feature names change often and may depend on your plan. If you don’t see an option exactly as described, look under Personalization in your settings.

Step 6: Test It With the Same Task Twice

Don’t assume your setup works. Prove it.

Pick a real task, like the intro for your next post, and run it twice: once in a plain chat with no instructions, and once with your style rules active. Then compare them side by side.

Look at the first sentence first. Openings show the difference fastest, because the model hasn’t had time to drift yet.

Then ask yourself:

  • Does the opening sound like something I’d actually write?
  • Does the argument build the way mine usually does?
  • Are the examples and comparisons the kind I’d use?
  • Are there any words I’d never say?
  • Is it copying phrases from my samples too literally?

If you answer “no” to most of these, your rules are still too vague. Go back to Step 3 and make them more specific.

When a draft misses, don’t just regenerate and hope. Tell ChatGPT exactly what’s wrong, then turn the fix into a permanent rule:

This is too formal. Compare it to my samples: I use shorter sentences,
I talk directly to the reader, and I’m more blunt. Rewrite it.
Then tell me which rule I should add to my Style Guide so this
doesn’t happen again.

Every correction makes your Style Guide sharper.

If a draft is completely off, change the input, not just the instructions. Try a different sample that’s closer to the task, or rewrite the part of your Style Guide that caused the problem. Sometimes one unusual sample is pulling the whole style in the wrong direction.

Why ChatGPT Stops Sounding Like You (and How to Fix It)

Have you noticed that the first draft sounds great, but by the fifth message it sounds like ChatGPT again? That’s called style drift, and it’s normal.

As a conversation grows, your samples and instructions sit further back in ChatGPT’s memory and lose their influence. The model slowly returns to its default voice: careful transitions, balanced phrasing, polished but generic sentences.

Here’s how to keep it on track:

  • Use a one-line anchor. Paste a short reminder at the start of each new task: “Follow my Style Guide. No em dashes. No fluff. Short sentences.”
  • Write section by section. Asking for a whole 2,000-word article at once almost guarantees drift. Ask for the intro, then each section.
  • Start fresh chats often. A new chat inside your Project, or with your Style File attached, reloads your style from the beginning.
  • Stick to familiar formats. If your samples are blog posts, expect weaker results when you ask for a sales contract or a speech.

What About Fine-Tuning?

You may have heard that you can “train” a custom AI model on your writing. This is called fine-tuning, and it’s real. But it requires hundreds of carefully formatted examples, technical skills to work with an API, and ongoing costs.

For bloggers, freelancers and content writers, it’s almost always overkill. The steps in this guide get you most of the way there for free.

Your 5-Minute Final Edit Checklist

Even with a perfect setup, ChatGPT gives you a draft, not a finished piece. Before you publish:

  • ☐ Read it out loud. Rewrite anything you’d never actually say.
  • ☐ Search for banned words with Ctrl+F (or Cmd+F).
  • ☐ Vary sentence length. Break up long sentences and combine choppy ones.
  • ☐ Add one thing only you know in each section: a real result, a client story or a strong opinion.
  • ☐ Rewrite the first and last lines yourself. Readers remember them most.

ChatGPT can learn how you write. It can’t know what you think, what you’ve lived through, or which jokes you’d cut. That part is still your job.

Final Thoughts

“Write like me” is not a magic prompt. Your voice lives in your actual writing: how you open, how long you stay on an idea, what you refuse to say. ChatGPT can only learn it from evidence.

Give it good samples, turn its analysis into concrete rules, save those rules where ChatGPT will always see them, and keep refining. After a few weeks, your first drafts will need a light edit instead of a full rewrite.

Start today: pick three pieces of your writing, run the analysis prompt from Step 2, and save your first Style File.

FAQ

Can ChatGPT really learn my writing style?

Yes, partly. With samples and clear rules, ChatGPT can match your tone, vocabulary, sentence length and formatting quite well. It’s weaker at deeper habits, like how you build an argument, and it can’t add your personal experience. Expect a strong first draft, not a finished piece.

How many writing samples should I give ChatGPT?

3 to 5 typical samples (about 1,000 to 2,000 words total) are enough for a noticeable improvement. For better results, use 5 to 10 samples for each format you write in, such as blog posts, emails and social media posts.

Does this method work with Claude or other AI tools?

Yes. The core method of collecting samples, analyzing your style and saving the rules works with any AI writing tool. The easiest way to move your voice between tools is a Style File: one document with your Style Guide and labeled samples that you attach at the start of a chat. Many tools also have their own version of custom instructions or projects where you can save it.

Are Custom Instructions enough to make ChatGPT write like me?

They’re a good start for basic preferences, like tone, banned words and formatting. Because the fields are short, they can’t hold your full Style Guide. For more depth, use a Project with your Style File uploaded, or ChatGPT’s writing style setting.

How do I stop ChatGPT from using em dashes?

Add a clear rule to your Custom Instructions, such as: “Never use em dashes (—). Use a period to start a new sentence or a comma to continue it.” If one slips through in a long chat, remind it with your one-line anchor.

Why does ChatGPT stop writing like me after a few messages?

In long chats, your instructions and samples lose influence as newer messages pile up, so ChatGPT drifts back to its default voice. Write section by section, repeat a short style reminder, and start fresh chats inside a Project or with your Style File attached.


📬 Want more practical prompts like these? Get AI writing tips in your inbox once a week. No spam.

Leave a comment

Your email address will not be published. Required fields are marked *