100% AI Text Detected is not AI text detected. Its pattern detected.
Let me prove it:
In this study I picked four AI models: Gemini, Grok, ChatGPT and Claude.
Procedure:
1) First, we will give this prompt to each AI:
Give me 200 word essay on smartphone usage in India.
2) We will detect AI Content using https://www.gptinf.com/detector
3) I have created some rules that AI need to follow to generate the text, we will upload those rules and give the below prompt to each AI.
follow these rules, Give me 200 word essay on smartphone usage in India.
4) We will detect AI Content again using the same website.
Gemini
Gemini latest version available for public on article published date was used, using fast mode.
Gemini without rules -> 0% Human-written, 100% AI-generated.
Gemini with rules -> 86% Human-written, 14% AI-generated.
Grok
Grok latest version available for public was used on the day article was published, using fast mode.
Grok without rules -> 0% Human-written, 100% AI-generated.
Grok with rules -> 84% Human-written, 16% AI-generated.
ChatGPT
ChatGPT latest version available for public was used on the day article was published.
ChatGPT without rules -> 0% Human-generated, 100% AI-generated.
ChatGPT with rules -> 74% Human-written, 26% AI-Generated.
Claude
Claude latest version available for public was used on the day article was published. Haiku 4.5 model was used.
Claude without rules-> 0% Human-written, 100% AI-generated.
Claude model. With rules -> 88% Human-written, 12% AI-generated.
And yes, for all of the above if I remove that red highlighted line, or write that much on my own, the result will look like this -
End Note
AI models trained on human-written text now generate text that resembles their training data — yet we flag this as “AI-generated.” The circularity is worth acknowledging: humans created the patterns, AI learned them, humans now mistrust those patterns when they reappear. Every text here will eventually train future models, making the distinction between original and generated operationally meaningless. Rather than detecting AI, we should ask what matters: whether ideas are substantive and thinking genuine. The advancement is real. What we choose to do with it — amplify or obscure human thinking — remains within our control.
So, in the comments, tell me how much of this article is human-written.
Connect me on Linkedin at https://www.linkedin.com/in/amritanshu-rawat/
Top comments (0)