This started randomly in our group. We were talking about AI detection and one of my friends said his own assignment got flagged. Everyone has their own thoughts and views about detection so we decided to test it properly instead of just guessing and try to find how to avoid plagiarism.
Everyone shared a few assignments a mix of fully written, AI assisted, and some where AI was used more heavily. Then we checked them across different detectors.
The surprise part for me was how inconsistent things were at first.
Two people with completely human-written work still got flagged higher than expected. But when we looked closely, those assignments were very perfect, clean structure, formal tone, no personal touch.
Then we tried something simple.
We took a few of those flagged assignments and rewrote small parts, added examples, changed wording to sound more natural, and checked again.
That’s where things got interesting.
Some tools did not change much, but when we tested again on Originality .ai, the score actually shifted in a way that matched the edits we made.
It felt less random and more like it was reacting to the writing style itself.
That kind of changed how I look at these tools.
Not something I did blindly trust, but definitely something that can help you understand how your writing comes across.
I did not expect a simple group test to be this eye opening.
Has anyone else tried something like this with friends or classmates?