Rewrite Machine-Sounding Text Without Inventing Facts
You get a rewrite that sounds like a person wrote it, with a clear list of what was changed. The locked fact list stops the model inventing statistics or stories to seem more human.
The prompt
Act as a careful native-speaker editor of {language} who specialises in removing the stiff, translated feel of AI-written text.
Task: Rewrite the draft below so it reads as natural {language} for the stated register, without adding anything the draft does not already say.
Register: {register}
Draft:
{draft_text}
Steps:
1. Mark the machine-like patterns: empty scene-setting openers, "not only X but also Y" constructions, stock phrases, passive chains, every sentence the same length, and a preachy final line.
2. List every fact, number, name and claim in the draft. These are locked: you may reword them but never add, remove or strengthen one.
3. Rebuild each sentence the way a native writer would say it, using direct active verbs, varied sentence lengths, and the conventions of the chosen register.
4. Check the rewrite against your fact list and delete anything invented, including anecdotes, statistics and causes.
Output:
First, the rewritten text only. Then a short list headed "Changes" with one line per pattern fixed, quoting the old phrase and the new one. Finish with one line confirming that no new facts were added.
Change the parts in curly braces. Use it on AI-drafted blog posts, product pages or emails in any language where the text reads as translated or generic.
What happened when we ran it
We sent
Act as a careful native-speaker editor of British English who specialises in removing the stiff, translated feel of AI-written text. Task: Rewrite the draft below so it reads as natural British English for the stated register, without adding anything the draft does not already say. Register: Friendly blog post for small business owners, conversational but clear Draft: In today's fast-paced business landscape, remote working is not only a passing trend but also a cornerstone of modern company culture, and its importance cannot be overstated. Research conducted by companies has shown that offering flexibility plays a crucial role in improving productivity. However, a holistic approach must be adopted in order to prevent communication breakdowns. In conclusion, organisations that wish to succeed in the future of work must adapt to this transformation. Steps: 1. Mark the machine-like patterns: empty scene-setting openers, "not only X but also Y" constructions, stock phrases, passive chains, every sentence the same length, and a preachy final line. 2. List every fact, number, name and claim in the draft. These are locked: you may reword them but never add, remove or strengthen one. 3. Rebuild each sentence the way a native writer would say it, using direct active verbs, varied sentence lengths, and the conventions of the chosen register. 4. Check the rewrite against your fact list and delete anything invented, including anecdotes, statistics and causes. Output: First, the rewritten text only. Then a short list headed "Changes" with one line per pattern fixed, quoting the old phrase and the new one. Finish with one line confirming that no new facts were added.
The model replied
Run on Claude Sonnet 5.5, 14 Nov 2026. Automatic check: 2 out of 5.
Mentioned for
Claude, ChatGPT, Gemini
Where it came from
Found on GitHub, shared by yucezerey: yucezerey/turkish-anti-ai-skill.
This is our own version of the idea, written and tested by us. The original may differ.