O que os professores devem fazer quando suspeitam de uso indevido de IA

Suspecting AI misuse is not the same as knowing it, and how a teacher responds shapes whether the outcome is fair. A sound process has four stages. Gather evidence first. Compare the work with the student's earlier writing and in-class samples. Check every citation, because fabricated references, such as real-looking journal articles that do not exist, are among the most concrete signs of unedited chatbot output. Look at version history if the work was written in a shared document. Note specifics: a sudden change in vocabulary, content that ignores the prompt's class-specific requirements, or claims the class never covered. Treat any AI-detector score as a weak signal at most; OpenAI withdrew its own classifier in 2023 for low accuracy, and detectors have documented false positives. Hold a private, non-accusatory conversation. Open with curiosity: ask the student to walk through how they wrote it, explain a key paragraph, or define a term they used. Students who wrote the work can usually talk about it; students who did not often struggle, and some will disclose. Avoid opening with 'I know you used ChatGPT', which invites denial and damages trust if you are wrong. Document as you go: dates, what you observed, the questions asked, what the student said and what was agreed. Follow your school's academic integrity policy, including any requirement to report. Choose proportionate consequences. Consider whether the assignment's AI rules were clear, whether this is a first incident, and how much of the work was affected. Options range from redoing the task under supervision, to a reduced grade, to a formal referral for repeated or serious cases. The aim is that the student ends up doing the learning. Consistency matters. Apply the same checks to every student, not only those whose writing style or language background makes them look suspicious.