This is a real problem. Even human-written poetry can get flagged if it's simple and clean. We're working on improving accuracy for our detector (AI Button) and open-sourcing our dataset so researchers can tackle false positives. What detector did you use? Would be great to test it against our dataset.
Loading…