In classrooms, AI detectors are easily used like verdict machines. Text detection can flag human writing, especially from language learners and translated work. The right use of a result is as grounds to start a conversation.
A result assesses how closely the statistics of a text resemble a generation model. It does not prove a student used AI, and it is worth knowing first that some human texts are flagged as AI.
Writing in a second language, machine-translated text, short pieces and formulaic templates produce more false alarms or come back inconclusive. Give results less weight for these assignments.
Drafts and revision history, an explanation of the writing process and a short oral check are far stronger grounds than a detection result. Revision traces in document files also tell part of that story.
State in the assignment how far AI may be used and how to disclose it. Clear rules reduce misunderstandings around results and make it easier for students to disclose their use.
This is general guidance, not legal advice. A scan result is a probabilistic assessment, not proof.