امسح بصمة ChatGPT من كتابتك: قواعد جاهزة من دليل ويكيبيدياRemove the ChatGPT Fingerprint From Your Writing: Ready Rules From Wikipedia's Guide

Key terms

AI writing fingerprint

Habits that repeat in AI text, like em dashes, lists of three and an "In conclusion" line. Readers spot them fast and stop reading.

Signs of AI writing

An advice page from Wikipedia's WikiProject AI Cleanup that collects common chatbot writing patterns with real examples.

Custom instructions

A ChatGPT setting where you write standing rules it applies to every new chat. In Claude the closest setting is personal preferences.

AI detector

A tool that tries to judge whether text was written by AI. Accuracy is poor: OpenAI shut down its own classifier in 2023 after it caught only 26% of AI written text.

The steps

1. 1

   Step 1

   Get the goal right first

   The goal is not to fool a detector or hand in work that isn't yours. The goal is a message that gets read to the end and sounds like you.

   If your school or job asks you to disclose AI use, disclose it. These rules improve style. They don't change honesty.

   Tip: Write your own ideas first and let the AI tidy them, not the other way around.

2. 2

   Step 2

   Learn the common tells

   The best known patterns from the Wikipedia guide and its summaries: em dashes instead of commas, "not just X, but Y", grouping everything in threes, inflated words like pivotal and transformative, too much bold and too many lists, and the summary ending like "In conclusion".

   Open an old ChatGPT reply and count the tells. You'll be surprised.

   PromptAsk it to find the tells in your text

   Copy

   Here is a text I wrote with AI help:
   [paste the text]
   
   List every AI writing tell in it: em dashes, "not just... but", lists of three, inflated words, extra bold, and endings that repeat the point. Quote the line for each one.

3. 3

   Step 3

   Have it read the Wikipedia guide itself

   If your ChatGPT has search on, have it open the page and pull the full list instead of relying on memory.

   en.wikipedia.org

   

   PromptPull the list from the source

   Copy

   Open the Wikipedia page "Wikipedia:Signs of AI writing" (en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing). List every writing pattern it describes, grouped by category, with one short example each. Skip patterns that only apply to Wikipedia articles.

4. 4

   Step 4

   Turn the list into writing rules

   In the same chat, ask it to turn the list into short, clear rules. One line each, written as never do X, do Y instead.

   Add a rule for your own voice, like the dialect or tone you actually use.

   PromptStanding rules

   Copy

   Turn that list into a strict set of writing rules, one line each, in the form "Never X. Instead, Y." Keep it under 1,200 characters so it fits in Custom instructions. Add these rules too: write in [your language or dialect], keep sentences short, no headings or bold in short messages, and end on the point without a summary.

5. 5

   Step 5

   Paste them into Custom instructions

   In ChatGPT open Settings, then Personalization, then Custom instructions, and paste the rules into the box about how ChatGPT should respond.

   In Claude paste them into the personal preferences box in Settings. Every new chat now starts with these rules.

   support.claude.com

   

   Tip: Rules apply to new chats. Start a fresh chat after saving.

6. 6

   Step 6

   Test it on a real message

   Take a long message ChatGPT wrote for you before and ask it to rewrite it. Compare the two versions and count the tells in each.

   If a tell comes back, add it to the rules by name.

   PromptRewrite

   Copy

   Rewrite this message with my rules, same meaning, shorter:
   [paste the message]
   Then tell me any rule you couldn't follow and why.

7. 7

   Step 7

   Read it yourself before you send

   The rules cut the tells. They don't remove all of them. Read the message out loud before you send it. If there's a line you'd never say in real life, change it.

   This is the last filter and the most important one.

Ready prompts to copy

Write [the idea] as a WhatsApp message to [the person or group]. In [your dialect], three sentences at most, no em dashes, no lists, no closing summary.

Read this post and answer yes or no: does it have any AI writing tell from my rules? If yes, fix it with the smallest change:
[paste the post]