For teachers
What to do when a paper is flagged
Five checks to make before you raise a flag with a student. The last one is a conversation, and it comes before any penalty.
The Human team · Sep 15, 2026
You have a paper in front of you and a result that says Mixed or AI. Before you say anything to the student, work through these five checks.
1. Check the length and the setting
Look at how long the text was. Under 50 words our detector does not score it, and from 50 to 149 words the reading is labelled weak. Then look at the setting. Sensitive (above 2%) flags at the first small sign of AI-written text, and it flagged 2 of 1,928 human-written documents in our tests. A flag at that setting is a prompt for a second look, not for a conversation about misconduct.
2. Read the marked passages
Open the passages our detector marked and read them next to the rest of the paper. One marked paragraph in a long paper is a different case from a paper marked from start to finish. Treat the marks as a reading. We have not measured how exactly they fall on the sentences an AI model wrote, so don’t quote them to a student as proof of which sentences those were.
3. Put it next to the student’s other work
Compare the paper with what the student has written for you before, in class and at home. Ask for drafts and version history if your school uses them. They show how a paper was written. A detector cannot.
4. Know what our detector has not been tested on
Every human-written document we tested was written before 2022. Writing from after AI tools became common has not been tested, and neither has writing done with grammar checkers or translation tools. If the student uses those, we cannot tell you how their writing will read.
If the student is learning English: our detector flagged none of 1,195 essays by English learners that were kept out of training as a fairness test. Those essays come from one university English language program, written between 2005 and 2012. That is a count, not proof of fairness.
A Human result does not clear a paper either. When an AI model lightly polishes a human paper, our detector mostly misses it. Of 39 polished papers, it flagged 3.
5. Talk to the student first
Ask the student how they wrote the paper. Tell them which tool flagged it and at which setting, and show them the passages. Whether an AI edit breaks a rule is your school’s policy, not the detector’s. Don’t make a flag the only reason for a penalty.
A flag from our detector is a reason to look closer. It is not proof that anyone cheated.
If your school is writing a policy on AI detectors and has questions about ours, write to hello@olive.is.
Our detector, run v1_8_s2_seed303. Settings and the polished-paper count measured 2026-09-15, source docs/reports/sensitivity_points_2026-09-14.md. Learner essays read on 2026-09-13 and recounted at all three settings on 2026-09-15, source docs/reports/v1_8_seed2_score_2026-09-13.md.