To use ChatGPT as a rewriter, paste your text and tell it four things: who will read it, what to change, what must stay exactly as it is, and how long the result should be. "Rewrite this" on its own gives ChatGPT nothing to aim at, so it rewrites toward its own default voice, and that's the voice people now recognise. Below is a template that works for any rewrite, copy-ready prompts for eight common jobs (shorter, more formal, friendlier, clearer, a different audience, your own voice, grammar only, a calmer tone), the reasons rewrites come out generic, and one email rewritten step by step.
How do you get ChatGPT to rewrite something for you?
Give ChatGPT your text inside clear markers and a short brief above it. Rewriting is one of the most common things people use it for: in OpenAI's study of how people use ChatGPT, writing made up about 40% of work-related messages, and about two-thirds of all writing messages asked ChatGPT to change text the user had provided, with the rest asking for new writing (Chatterji et al., NBER Working Paper 34255). "Editing or Critiquing Provided Text" was the most frequent writing request of all.
This template covers most rewrites. Fill in the brackets:
Rewrite the text in the <text> tags.
Reader: [who will read it, and what they need from it]
Change: [shorter / more formal / friendlier / clearer]
Keep: [names, dates, numbers, phrases that must not change]
Length: [a word count, e.g. under 120 words]
Return only the rewritten text.
<text>
[paste your text here]
</text>
The tags matter more than they look. OpenAI's prompting guide notes that "XML tags can help delineate where one piece of content (like a supporting document used for reference) begins and ends" (OpenAI prompt engineering guide). Without them, ChatGPT sometimes treats a line in your email as an instruction. The last line stops it opening with "Sure! Here's a more polished version."
Why do ChatGPT rewrites sound like ChatGPT?
When you don't name a reader or a voice, ChatGPT picks the safest one: even, polished, slightly formal, and fond of the same words. The habits are measurable. A study of over 15 million biomedical abstracts found that at least 13.5% of 2024 abstracts were processed with language models, detectable through a jump in certain style words; "delves" appeared 28 times more often than the pre-2024 trend predicted (Kobak et al., Science Advances 2025, arXiv:2406.07016). Readers have learned those words too, which is why a ChatGPT rewrite can make an honest email feel automated.
The other cause is inflation. Asked to "make it better," the model tends to add a greeting, a compliment, a summary line and a sign-off you didn't write, so a three-line note comes back as eight. OpenAI's own guidance points at the fix: say the tone with a plain adjective ("formal, informal, friendly, professional, humorous, or serious") and give a length in numbers, because "a 3 to 5 sentence paragraph" is clearer than "fairly short" (OpenAI Help Center; OpenAI API best practices).
How do you make a ChatGPT rewrite sound human?
Make it sound like you specifically. These four habits do most of the work:
- Give it a sample of your writing. Paste 150 to 300 words you wrote for a similar reader, ideally two different pieces, and ask it to match the sentence length, word choice and punctuation. OpenAI's guide recommends showing "a diverse range of possible inputs with the desired outputs" when you give examples.
- Name what to keep. Your opening line, a phrase you like, contractions, the way you sign off. Anything you don't name is fair game for the rewrite.
- Ban specific words by name. "Don't sound like AI" is too vague to follow. "Don't use delve, seamless, elevate, leverage or 'I hope this finds you well'" is a rule it can check. Pair it with what to do instead, such as "use short sentences and plain verbs."
- Ask for the smallest change that fixes the problem. "Change as little as possible" keeps your structure and most of your words.
Copy-ready rewrite prompts for 8 common jobs
Each prompt below assumes you paste your text between <text> and </text> tags at the end, as in the template above. Replace anything in brackets.
1. Make it shorter
Give a word count. "Shorter" alone often trims a sentence or two.
Shorten the text in the <text> tags to under [80] words for [a manager reading on their phone].
Keep every date, number and name, and keep the request at the end.
Cut background they already know. Return only the shortened text.
2. Make it more formal
Useful for clients, complaints and applications. Stop it from adding ceremony.
Rewrite the text in the <text> tags in a formal, professional tone for [a client I haven't met].
Keep the meaning, every fact and the order of points. No slang, no contractions.
Don't add greetings, compliments or closing lines that aren't in the original.
Keep it about the same length.
3. Make it friendlier
Friendly usually means shorter and plainer. Say what friendly shouldn't include.
Rewrite the text in the <text> tags so it sounds warm and friendly, for [a colleague I work with every day].
Keep the request clear and near the top. Use contractions and plain words.
No exclamation marks, no emojis, and no "I hope you're well."
4. Make it clearer
Name the reader who's currently confused. That decides which words count as jargon.
Rewrite the text in the <text> tags so [a new customer with no technical background] understands it on first read.
Put the main point in the first sentence. Replace jargon with plain words, or explain a term briefly the first time.
Keep sentences under 20 words where you can. Keep all facts and numbers exactly as written.
5. Rewrite it for a different audience
Say who it was written for, who it's for now, and what the new reader needs to do with it.
The text in the <text> tags was written for [engineers]. Rewrite it for [the sales team],
who need to know [what changed for customers and what to tell them].
Keep the facts. Drop details this audience won't use. Length: about [150] words.
6. Keep my voice
The one to use for anything that goes out under your name.
Here are two samples of my writing. Match their sentence length, word choice and punctuation.
<sample>[paste 150 to 300 words you wrote]</sample>
<sample>[paste a second piece]</sample>
Rewrite the text in the <text> tags so it sounds like me. Change as little as possible:
fix what's unclear and keep my phrases where they work.
Don't use these words: [delve, seamless, elevate, leverage, robust].
Don't add dashes, bullet points or headings that my samples don't use.
7. Fix grammar only
Without the limits, ChatGPT tends to "improve" sentences that were already fine.
Fix only spelling, grammar and punctuation in the text in the <text> tags.
Don't change word choice, tone, sentence order or length. Leave correct sentences alone.
Return the corrected text, then a list of every change you made.
8. Change the tone, keep the message
For the email you wrote while annoyed. It keeps the complaint and drops the heat.
I wrote the text in the <text> tags while frustrated. Rewrite it so it's calm and direct, for [my manager].
Keep every point I make, including the complaint. Remove sarcasm and blame.
Don't apologise for anything I didn't do. Keep it under [150] words.
Worked example: rewriting an email with ChatGPT
Here is a typical first draft, the kind written in a hurry between meetings:
Hi Sam, so sorry about this but just wanted to let you know that the report is probably going to be late, we had some issues with the data from the vendor and I've been chasing them since last week but they haven't come back yet so realistically it might be next Wednesday instead of Friday, hopefully that's ok? Let me know if that's a problem and we can figure something out. Thanks so much!
A bare "make this more professional" tends to return something longer and stiffer, with a greeting and a closing paragraph you'd never write. This prompt fills in the template instead:
Rewrite the email in the <text> tags.
Reader: Sam, a client who expects the report this Friday.
Change: shorter, clearer and more confident. Keep one short apology.
Keep: the new date (next Wednesday), the reason (vendor data is late), the offer to work something out.
Length: under 80 words. Keep "Hi Sam" and "Thanks".
Don't add "I hope this email finds you well" or any line I didn't write.
Return only the email.
A rewrite that follows that brief looks like this:
Hi Sam,
The report will be late. Our data vendor still hasn't sent the figures we've been chasing since last week, so I now expect to deliver it next Wednesday instead of Friday. Sorry for the change.
If Wednesday causes a problem on your side, let me know and we'll work something out.
Thanks,
It went from 75 words to 54. Every fact from the draft is still there: the delay, the reason, the new date, the apology, the offer. What changed is the order (the news comes first) and the confidence ("I now expect" in place of "probably" and "realistically it might"). It still sounds like a person wrote it, because the brief said what to keep and what not to add.
Want to rewrite a prompt instead?
Rewriting your text and rewriting your prompt are different jobs. If what you want is a better answer from ChatGPT, the thing to rewrite is the instruction, and our guide to rewriting ChatGPT prompts has the manual steps and a one-click method.
PrePrompt is a chrome extension that adds a Rewrite button to ChatGPT, Claude, Gemini and Perplexity. It launches soon with 30 free rewrites a month, no card required; join the waitlist to hear the day it goes live. It rewrites the prompt in the box before you send it, and keeps exact text you wrote (file names, numbers, quotes, links) as written. It's built for prompts, so for editing an email itself, the prompts above are the tool.
TL;DR
- To make ChatGPT rewrite your text, paste it between tags and say who reads it, what to change, what to keep, and a word count.
- Rewriting is a core use: about two-thirds of ChatGPT writing messages ask it to change text the user provided.
- Rewrites sound generic because the model defaults to its own voice and vocabulary; "delves" showed up 28 times more often than expected in 2024 research abstracts.
- To keep your voice, paste a sample of your writing, name what to keep, ban specific words, and ask for the smallest change.
- For grammar-only fixes, say so explicitly and ask for a list of changes.
About the author. Tanuj writes PrePrompt's how-to coverage and works on the product's marketing. PrePrompt builds and tests prompt rewriting across ChatGPT, Claude, Gemini and Perplexity, which informs the advice in this post.
Sources: How People Use ChatGPT (Chatterji et al., NBER Working Paper 34255, 2025); Delving into LLM-assisted writing in biomedical publications through excess vocabulary (Kobak et al., Science Advances, 2025, arXiv:2406.07016); Prompt engineering (OpenAI API docs); Prompt engineering best practices for ChatGPT (OpenAI Help Center); Best practices for prompt engineering with the OpenAI API (OpenAI Help Center).
Related reading: how to rewrite a ChatGPT prompt automatically, what a good prompt looks like and why ChatGPT gives vague answers.












