Nowadays, we consume a lot of AI-generated content, as it is being used everywhere, from newspapers to textbooks. As a result, we humans are also starting to write in a way that resembles AI-generated content. The irony is that, in recent times, even content written entirely by humans can be flagged as AI-generated when passed through an AI detector.
I reviewed a paper recently and got totally stuck on page one. I had to ask ChatGPT to explain the sentences. I realized the authors were using hard concepts on page one that they didn't explain until page six.
That was my lightbulb moment. The paper wasn't badly written for its reader. Its reader just wasn't me.
We are stuck in a bad loop right now:
Humans read in order. We need the basics explained first. But an AI reads the whole document at once. It does not care if a word is used five pages before it gets explained.
I got tired of reading these messy papers. I built a free, open-source skill to break the loop. It forces AI models (like Claude, ChatGPT, and Cursor) to pass a "first-page test". They have to explain terms in order, keep things simple, and stop sounding like a robot.
If you or your lab uses AI to write or review papers, you can get the skill file here: https://github.com/Aadarshttech/ai-research-paper-humanizer
It really helps make papers readable by normal people again. Let me know if anyone else is dealing with this same headache!