Professors read long, technical writing where one generated section can hide inside otherwise honest work. Winston AI highlights the sentences that read as machine-written so you can focus your review, then fold that evidence into your department's process. Below: handling long documents, what due process should look like, and where scores fall short.
Due process matters more than the number
An integrity case can affect a student's degree, visa or funding, so the bar is high. A detection score should prompt a closer reading and a meeting. It should never be the only item in a case file. Write down what you saw, what the student said and what other evidence you looked at.
- Say in the syllabus whether AI tools are allowed, and for which tasks.
- Scan the same way across every section of a course, so no group gets more scrutiny than another.
- Give the student a chance to explain and to share drafts, notes or version history before any finding.
- Keep the PDF report with your case notes, not in place of them.
Where academic writing trips detectors
Formal academic prose is formulaic by design. Abstracts, methods sections and the stock phrasing of a field can read as machine-like even when a person wrote every word. Students who learned English from careful textbook models can see the same effect. Read what an AI detection score means before you weigh a result, and give most attention to the sections where original thinking should appear.
For theses and capstone projects, the best protection is built into supervision. Ask for chapters as they're written, keep your comments, and scan drafts at the same points for every student. A flagged final chapter can then be compared with drafts you've already seen, and a student whose voice has stayed consistent has the record to show it.