Section and peer reviews seems doable. But how can AI text be detected when I. One sentence a minor manual change is made? And what about false positives? How will false positives handled? A common problem, also in plagiarism detection. Which often fails in subtle ways.