How to Make ChatGPT Write Like You: Every Method in 2026

To make ChatGPT write like you, give it your own writing and clear rules in the places it reads every time: a style profile in custom instructions, a Project with your best samples as files, and Memory for the details it should keep. In 2026 you can also turn on Writing Style in ChatGPT Work, which learns from your connected apps. None of these retrain the model, so test the output before you trust it.

ChatGPT has more ways to learn your writing than it did a year ago, and fewer than most guides claim. Some settings changed this year, one is brand new, and one is being retired. Here's every method, what OpenAI says about each, and a setup that works in about 20 minutes.

First, what "learn" means here

When people ask whether ChatGPT can learn their voice, they usually picture it studying their writing and getting better each week. That isn't how it works.

The model behind ChatGPT doesn't change when you use it. What changes is what it reads before it answers: your instructions, your saved memories, the files in a Project, your past chats. Every method below is a way of putting the right examples in front of it.

That's good news and bad news:

• Good: you control exactly what it learns from. Better samples, better output.

• Bad: it imitates; it doesn't measure. Nothing tells you how close the draft is to you, so you have to check.

Every method at a glance

Limits from OpenAI's help pages, checked 5 November 2026.

Method 1: paste samples into a chat

The fastest method, and the one OpenAI itself teaches. Its Academy guide says to "paste in examples of your writing and ask: 'Summarize my tone and style…'".

Do it like this:

1. Pick 5 to 10 pieces of the same type you want it to write. LinkedIn posts for LinkedIn posts, emails for emails. Mixing types teaches it an average of both.

2. Pick your best pieces, not your most recent. It copies whatever you give it, weak lines included.

3. Ask it to describe your style before it writes anything: how you open, how long your sentences run, words you use often, words you never use, how you end.

4. Correct the description. This is the step people skip, and it matters most.

The catch: it lasts one chat. Start a new one and it's gone. Use this to build your style profile, then move the profile somewhere permanent.

Method 2: custom instructions

Custom instructions apply to every new chat. You'll find them under Settings, then Personalization. OpenAI's limit is 1,500 characters on Free and Go, and 5,000 on Plus, Pro, Business, Enterprise and Education (paid plans went up from 1,500 in July 2026). Any guide that still says "1,500 per field" is out of date.

What to put in them:

• The style profile from Method 1, edited down to rules: "Short sentences. Open with the point. No exclamation marks."

• A short list of words you never use.

• Who you write for, in one line.

Keep it to rules, not samples. Samples eat the space fast, and they belong in a Project.

Method 3: base style, tone and characteristics

Under Personalization you'll also find Base style and tone, with seven presets: Default, Professional, Friendly, Candid, Quirky, Efficient and Cynical (Nerdy was retired in March 2026). Next to it are characteristics: warmth, enthusiasm, headers and lists, and emoji, each set to more or less.

These change how ChatGPT talks to you. They don't decide how your drafts read. OpenAI says so directly: when you ask for "an email draft" or a "social media post", ChatGPT "will match the tone and style to your instructions and the context."

Two settings worth changing anyway: turn headers and lists and emoji down if you never use them. It stops them leaking into drafts.

Method 4: Memory

Memory lets ChatGPT keep details across chats. You can ask it to remember something directly ("Saved memories are details that you explicitly ask ChatGPT to remember"), and it can also reference your past chats. OpenAI says there is "no separate storage limit" for what it can reference through chat history, and that memory "can vary by plan, region, platform, and workspace settings."

Use it for small, stable facts: "I never use emoji in posts." "I write in British English." "My audience is first-time founders."

Don't rely on it for your whole style. Memory decides for itself what's relevant, and you can't see when it's being used. Your style rules belong in instructions you can read.

Method 5: Projects (the best place for your writing)

A Project is a workspace with its own instructions, files and, optionally, its own memory. For writing, it's the strongest option most people have:

• Project instructions "apply only within that project and override your global custom instructions." OpenAI publishes no length limit for them.

• Files: 5 per project on Free, 25 on Go and Plus, 40 on Edu, Pro, Business and Enterprise (up to 10 at once).

• Project-only memory: set it and "your previously saved memories are not referenced in project chats." Useful when you write for clients or for more than one platform.

Make one Project per kind of writing: "LinkedIn posts", "Newsletter", "Client emails". Put your style profile in the instructions and your best 10 to 25 pieces in the files.

Method 6: Writing Style in ChatGPT Work (new in September 2026)

This is the first ChatGPT feature built to learn your writing directly. OpenAI's own wording: "In ChatGPT Work on the web, writing style uses your own writing from connected apps to help drafts sound like you."

What we could confirm on OpenAI's pages:

• ChatGPT's launch post (7 September 2026) named Gmail, Google Drive, Slack and SharePoint, and said it "will learn your writing style from your emails, messages, and files."

• You connect the apps, then switch on Reference my writing style. "Connecting an app alone doesn't complete setup."

• To turn it off: Settings, Personalization, Writing, on the web.

• ChatGPT Work rolled out to paid plans except Free and Go.

Two honest limits:

• It learns your email voice. Your emails, Slack messages and docs are not how you write a LinkedIn post or a tweet. A post written from your email voice reads like an email.

• It still reads like AI to detectors. In PCWorld's hands-on test, it copied the author's em dashes and pet phrases, and GPTZero still flagged the text as AI. That's one test, but it's the only published one so far.

Method 7: custom GPTs (don't start one now)

For two years, the standard advice was to build a custom GPT with your writing as knowledge files. That advice has expired, according to OpenAI's retirement FAQ:

• "Custom GPTs are scheduled to retire on Dec 11, 2026."

• "New GPT creation and publishing are not available on personal ChatGPT accounts, including Free, Go, Plus, and Pro."

• GPTs never used your memory anyway: "GPTs do not use saved memory, custom instructions, or previous conversations."

If you have one, OpenAI says its instructions will become a skill in a plugin. Skills are generally available on Enterprise and Edu plans. On a personal plan, move your GPT's instructions and files into a Project.

How much ChatGPT can read at once

Your samples compete for space with everything else ChatGPT reads. OpenAI's pricing page lists these context windows:

OpenAI notes that the usable space "is smaller than the total window", because instructions and memories take room too. In practice: 10 strong samples beat 50 average ones.

Source: OpenAI help pages and chatgpt.com/pricing, 5 November 2026.

A 20-minute setup that works

1. Collect 10 pieces of one type of writing you're proud of.

2. Profile: paste 5 into a new chat and ask for your style: openers, sentence length, favourite words, banned words, endings. Correct it.

3. Rules: put the corrected profile in custom instructions (or project instructions), written as short rules.

4. Project: create a Project for that type of writing and add all 10 pieces as files.

5. Memory: save two or three stable facts ("no emoji", "British English").

6. Test with the three checks below before you post anything.

For LinkedIn-specific prompts and a banned-word list, see our guide to making ChatGPT write LinkedIn posts like you.

Before and after

An illustrative example, written for this guide. The ask: "Write a LinkedIn post about hiring our first salesperson."

Illustrative example, written for this guide.

No setup:

Excited to share a big milestone! 🚀 We've just welcomed our first sales hire to the team. This marks an incredible new chapter in our journey, and we couldn't be more thrilled about what's ahead. Here are 3 lessons we learned along the way…

After a style profile and 10 samples in a Project:

We waited too long to hire our first salesperson.

For 14 months I did every demo myself. I told myself it was "founder-led sales." It was really me not wanting to let go.

She closed more in her first month than I did in my last three.

Same facts. The second one has an opening that makes a point, a real number, and an admission. That's what a style profile does: it removes the default, so what's left can be yours.

How to test whether it writes like you

Run these three checks on every new setup:

Run it on every new setup.

1. The blind pick. Write one piece yourself and get ChatGPT to write two on the same topic. Ask someone who reads your work to pick yours. If they can't, it's working.

2. Your tells. Check your own patterns: how you open, your sentence length, the words you'd never use. One miss is a rule to add.

3. The AI tells. Look for the patterns that survive almost any setup: em dashes you don't use, "It's not X, it's Y", three-item lists, a neat moral at the end. Add each one you find to your rules.

For a fuller checklist, see how to tell if an AI writes like you.

Where ChatGPT's methods stop

ChatGPT's methods are worth setting up. They also share three limits:

• They imitate, they don't measure. Nothing scores a draft against your real writing. You're the only check.

• One voice for everything. Unless you split your writing into Projects, your LinkedIn posts, tweets and emails all pull from the same rules.

• They forget what you post. Your best new posts don't flow back in. You have to update the files yourself.

Where VoiceMoat fits

VoiceMoat is an AI social media tool for LinkedIn and Twitter, and voice is the whole point of it. audenAI, the brain inside VoiceMoat, builds your Voice DNA from up to 200 of your own published posts, keeps a separate voice for each platform, and gives every draft a Voice Match score before you post it. When you publish more, it learns from those posts too.

If you mostly write emails and docs, ChatGPT's Writing Style may be all you need. If your writing lives on LinkedIn and Twitter, try VoiceMoat free and compare the drafts side by side.