[[{“value”:”
Hey AI Breakers 👋
There’s more AI-written content out there than ever.
And now there are tools built to catch it. They flag the same patterns every time: the punctuation, the phrasing, the tone that gives ChatGPT away.
Today we’ll show you a way to get ChatGPT to write copy that reads like a human wrote it and gets past AI detection tools. You set it up once, and it applies to every chat after that.
Let’s go 👇
🕵️ Step 1 → Make ChatGPT Expose Its Own Tells
There’s a public, constantly updated list of everything that gives AI writing away. Instead of guessing what “sounds robotic,” send ChatGPT to go read it.
✅ Use this to get the complete list of tells in one place, straight from the source.
Prompt:
You are an editor who specializes in spotting AI-generated writing. Go to Wikipedia and read the page titled "Wikipedia:Signs of AI writing" (https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing). 1. List every single sign of AI writing the page describes. Don't summarize or skip any. Include punctuation habits (like em dashes), overused words and phrases, sentence structures, tone, formatting habits, and filler. 2. Group them into categories: Punctuation, Vocabulary, Sentence Structure, Tone, Formatting, Filler & Disclaimers. 3. For each sign, give one short example of what it looks like in a sentence. Output as a clean numbered list. No commentary, no introduction, no closing summary.
💡 Tip: If ChatGPT says it can’t browse, click the + in the message box, turn on Web Search, and send the prompt again. You should get 30 to 50 signs back. If it comes back with 10, tell it “That’s not the full page. Go back and list all of them.”
📜 Step 2 → Turn the Tells Into Rules It Can’t Break
A list of tells is interesting. A rulebook it has to follow is what actually changes the output. Stay in the same chat so it still has the full list in front of it.
✅ Use this to get a compressed instruction prompt that fits ChatGPT’s settings and covers every sign.
Prompt:
You are a writing coach who trains people to write plainly and sound like themselves. Using the full list of AI writing signs above, do the following: 1. Turn every sign into a strict writing rule you must never break. Write each rule as a direct command ("Never...", "Always..."). 2. Merge overlapping rules so there are no duplicates. Order them from the most obvious tell to the least obvious. 3. Add 5 rules for what good human writing DOES do: varied sentence length, contractions, specific details instead of vague claims, plain everyday words, and committing to a clear opinion. 4. Compress all of it into a permanent instruction prompt I can paste into your custom instructions. It must be under 1,400 characters, written as commands to you, and start with: "Follow these writing rules in every response, without exception:" Show me the full rulebook first, then the compressed instruction prompt in a separate code block.
💡 Tip: The 1,400-character limit matters. ChatGPT’s custom instructions box caps at 1,500, and you want a little room to add your own rules later. If the output is too long, reply “Cut it to 1,400 characters, drop the least important rules first.”
⚙️ Step 3 → Install It (So It Never Forgets)
This is the part most people skip. Pasting rules into a chat works once. Pasting them into settings works forever.
Here is what you need to do:
“}]] Read More in The AI Break
