I enjoy sharing my findings on Kaggle and Medium. Lately, I keep getting comments asking why is my human-written content flagged as AI? I spend hours on my Python code and the explanations. Is it because I use a lot of mathematical terms and structured data definitions?
3 answers
Mathematics and code explanations are the "bread and butter" of AI training sets. When you explain a concept like Gradient Descent, you likely use the same logical flow as a model would. To prevent being flagged, focus on the "intuition" rather than just the formulas. Use unique visualizations or describe a specific time your model failed in an unexpected way. These "edge case" stories are hard for AI to fabricate convincingly. Also, avoid using standard AI-generated summaries at the top of your posts, as those set a bad tone for the detector.
Do you find that your writing style changes significantly when you are switching between writing code comments and writing the actual blog narrative?
Try recording yourself speaking the explanation and then transcribing it. Spoken language has natural "imperfections" and rhythms that AI tools rarely replicate perfectly.
That is a brilliant suggestion, Diana. Transcribing a natural conversation or a lecture usually yields a much more "human" score than typing directly into a text editor.
Definitely. My code comments are very brief, but my blog prose is much more formal. I think the contrast might be confusing the detectors, or perhaps my formal prose is just too close to the academic style that AI mimics. I should try to make my blog voice more conversational and less like a thesis.