Eight copy-and-paste prompts for ChatGPT, Claude and Gemini that make AI text sound more human, an honest look at where prompting stops working, and a humanizer for the part prompts cannot fix.
By the Pure Write TeamUpdated October 2026
No card required. Free to try.
Paste your prompted output
Turn AI drafts into clear, natural writing
Ready
Refine mode keeps your meaning but improves tone, flow, and clarity.
Paste your AI-generated or rough draft here…
💡 Tip: Press Ctrl+V to paste or drag & drop a .txt file
0 words
0 words·Quality: not checked yet
We never store or train on your text
Built to reduce AI-detection flags
See your live AI-detection score
How it works
Prompt first, then finish it here
1
Run a prompt below
Copy the prompt that fits your job, paste it into ChatGPT, Claude or Gemini with your text, and get a cleaner draft.
2
Paste the result here
Drop the prompted output into the editor above. PureWrite rewrites the rhythm and word patterns the model cannot see in its own writing.
3
Check the score
Read the built-in AI score on the result, then do a final read for meaning before you use it.
8 prompts to humanize AI text
Each prompt is written for one job. They work in ChatGPT, Claude and Gemini. Replace the bracketed parts, and paste your text at the end. The examples show the kind of change to expect; your output will differ.
Prompt 1
General rewrite
Use it when: You have a full AI draft and want a cleaner, more natural version before you edit it yourself.
Rewrite the text below so it reads like a thoughtful person wrote it, not an AI.
Rules:
- Keep every fact, name, number and claim. Do not add new information.
- Vary sentence length on purpose: some short, some long, a few medium.
- Prefer plain, specific words over abstract ones.
- Cut filler openers like "In today's world" and closers like "In conclusion".
- No em dashes, no bullet points, no headings.
Return only the rewritten text.
TEXT:
[paste here]
Before
In today's fast-paced world, effective time management is essential for students to achieve academic success.
After the prompt
Most students do not run out of ability. They run out of hours, and the ones who plan their week tend to get more done.
Prompt 2
Academic register
Use it when: An essay or report has to stay formal, but the AI version sounds like a brochure.
Revise the passage below for a university assignment.
- Keep a formal academic tone, but write like a strong undergraduate, not a press release.
- Keep the thesis and every citation exactly as written, including (Author, Year) brackets.
- Replace vague phrases ("plays a crucial role", "a wide range of") with the specific claim.
- Mix sentence lengths. Allow one short, direct sentence per paragraph.
- Do not use: delve, tapestry, multifaceted, pivotal, underscore, landscape.
Return only the revised passage.
PASSAGE:
[paste here]
Before
Social media plays a crucial role in shaping a wide range of political attitudes among young people (Smith, 2021).
After the prompt
Young voters now meet most political arguments in a feed rather than a newspaper, and that shift changes which views they encounter first (Smith, 2021).
Prompt 3
Discussion post reply
Use it when: You need a reply to a classmate on Canvas or Blackboard that sounds like a person in the class.
Write a reply to my classmate's discussion post below.
- 120 to 180 words, first person, conversational but respectful.
- Start by engaging with one specific point they made, not with "Great post!".
- Add one point of my own: [your point or a reading from class].
- End with a genuine question for them.
- No headings, no lists, no em dashes.
CLASSMATE'S POST:
[paste here]
Prompt 4
Reflection paper
Use it when: The assignment asks what you learned or felt, and the AI draft sounds like nobody was there.
Help me turn my notes into a reflection paragraph.
- Use first person and past tense for what happened, present tense for what I think now.
- Only use the experiences in my notes. Do not invent events, feelings or people.
- Include one moment where I was wrong or uncertain.
- Plain language. No "This experience taught me valuable lessons".
MY NOTES:
[paste your bullet notes here]
Prompt 5
Match my writing sample
Use it when: You want the output to sound like you specifically, not like a generic human.
Here is a sample of my own writing (about 200 words):
[paste your writing]
Study its sentence length, vocabulary, punctuation habits and how formal it is. Then rewrite the text below in that same voice. Keep all facts. Do not make it more polished than my sample.
TEXT TO REWRITE:
[paste here]
Prompt 6
Kill the AI vocabulary
Use it when: The draft is fine structurally but is full of words that make readers and detectors suspicious.
Edit the text below and replace any of these words or phrases with plainer, more specific wording: delve, tapestry, testament, multifaceted, pivotal, crucial, underscore, landscape, realm, navigate, foster, seamless, robust, leverage, "it's worth noting", "in today's world", "plays a vital role".
Change nothing else. Return the edited text only.
TEXT:
[paste here]
Prompt 7
Vary sentence length
Use it when: Every sentence is roughly the same length, which is the rhythm detectors notice first.
Rewrite the text below with deliberately uneven sentence lengths.
- At least one sentence under 6 words per paragraph.
- At least one sentence over 25 words per paragraph.
- Do not start two consecutive sentences the same way.
Keep the meaning identical. Return only the text.
TEXT:
[paste here]
Prompt 8
Remove em dashes and tidy triplets
Use it when: The draft leans on em dashes and lists of exactly three things, two strong AI habits.
Edit the text below:
1. Remove every em dash. Use a period, a comma or parentheses instead.
2. Wherever three items are listed in a row ("clear, concise, and compelling"), keep the one or two that matter or rewrite the sentence.
3. Remove "not X, but Y" constructions unless the contrast is essential.
Return the edited text only.
TEXT:
[paste here]
Prompting genuinely helps. It removes the obvious tells: stock phrases, em dashes, listicle structure, the closing summary. Human readers notice those first, so a good prompt makes text better to read.
Problem
Prompt fixes it?
Why
Stock AI words and phrases
Mostly
Explicit word bans work well, though models drift back to them in long outputs
Em dashes, triplets, bullet habits
Mostly
Concrete formatting rules are easy for a model to follow
Sounding like you
Partly
A writing sample helps, but the model smooths your quirks out
Uniform sentence rhythm
Partly
Models follow length rules for a few paragraphs, then regress to an even beat
Predictable word choice (low perplexity)
Rarely
The model still picks the words it finds most likely; that is how it works
A rule of thumb from how language models generate text, not a benchmark. Model behavior changes between versions, so test on your own text.
Why prompts plateau
Detectors do not look for the word "delve". They look at how predictable each word is given the words before it. When you ask ChatGPT to write "like a human", the same model is still choosing the next word, and it still prefers the likely choice. The surface changes; the statistical fingerprint mostly does not. A model cannot fully un-model itself.
That is the gap a dedicated humanizer fills. PureWrite rewrites at the sentence and paragraph level with the specific goal of breaking that even, predictable texture, and then shows you an AI score on the result so you are not guessing. It is designed to pass Turnitin, GPTZero, ZeroGPT, Originality.ai and Copyleaks, though no tool can promise how a detector will score a particular text.
Want to see it on your own text? Paste a prompted output into the editor at the top of this page, or check it first with the AI detector.
Prompt tweaks for ChatGPT, Claude and Gemini
Each model has its own habits, so the same prompt needs a different nudge.
Model
Typical tells
Add this line
ChatGPT
Upbeat filler, "In today's world" openers, neat summaries at the end of every paragraph.
Add: "Do not summarize at the end. Stop when the point is made."
Claude
Hedging ("it's worth noting"), balanced both-sides conclusions, long dashes for asides.
Add: "Take a clear position. Do not hedge or present both sides unless asked."
Gemini
Bullets and bolded headers even when you asked for prose; generic transitions.
Add: "Write continuous paragraphs only. No lists, no bold text, no headers."
More on each model: the ChatGPT humanizer and the Claude humanizer, which also compares the popular Claude "humanizer skill" with a dedicated tool.
A workflow that holds up
1
Write your own outline
Your thesis and points, in your words, before any AI is involved.
2
Draft with your prompt of choice
Prompt 1 or 2 for most writing, prompt 5 if you have a writing sample.
3
Humanize
Paste the output into PureWrite to fix the rhythm and word patterns the prompt left behind.
4
Add specifics
Examples, sources and details only you know. This is what makes writing yours.
Using this for school?
Check your course policy first; plenty of instructors allow AI for brainstorming but not drafting. Our guide to humanizing AI text covers the editing side, and the Turnitin AI humanizer page explains what Turnitin shows your instructor.
Copy the prompts freely
Every prompt on this page is free to use, edit and share. If a prompt gets you most of the way, the PureWrite plans are there for the rest.
No sign up needed to copy a prompt
Try the humanizer on a short passage with a free account, no card
PW
Pure Write Team
Updated October 2026
Why trust this guide
We build both sides of this problem: a humanizer that rewrites AI text and an AI detector that scores it. That means every guide here is written by the same team that tunes the engine against the patterns AI detectors flag, so the advice reflects what actually passes today, not last year.
FAQ
Questions, answered
For most writing, a prompt that bans stock AI phrases, forces uneven sentence lengths and forbids em dashes and bullet points does the most good. If you have a sample of your own writing, the match-my-voice prompt is better still. Prompt 1 on this page combines the general rules.