With AI tools within reach of every student, teachers need practical ways to judge whether an assignment reflects what the student actually knows. At the same time, students who write on their own want to make sure a well-written text is not mistaken for a generated one. This guide sets out an approach that works for both sides: one based on evidence of the process, on conversation and on clear rules.
1. Evidence of the writing process
A finished text shows little about how it was made. The process shows much more. Along with the final assignment, teachers can ask for:
- Drafts and intermediate versions: an outline, a first draft with comments, the revised version.
- Version history: online document editors keep a record of changes that shows the text being built step by step, with cuts, additions and corrections.
- Reading notes and a list of sources consulted: they show the path from research to text.
Tip for students: write in an editor that keeps version history, save your drafts and hold on to your notes. They are your best defense if questions ever come up about your work.
2. A conversation about the text
A short conversation is often more revealing than any analysis of the text. The teacher can ask why the student chose a particular argument, what they meant in a paragraph, where a piece of information came from or how they would explain the main idea in other words. Students who wrote the assignment usually answer naturally, even if they do not remember every sentence.
The tone should be curious, not interrogating. The goal is to understand the work, not to catch someone in a contradiction.
3. Assignments designed to show what students know
Some assignment designs make misuse of AI less useful and the student’s own work more visible:
- Ask students to connect the topic to class discussions, to a reading done in the course or to their own experience.
- Split the work into stages with partial submissions: topic, outline, first draft, final version.
- Include a part written in class, by hand or on a computer, that serves as a reference for the student’s style.
- Ask for a short reflection on the process: what was hardest, what changed between versions, which sources helped.
- State in writing which uses of AI are allowed for that assignment and how they should be declared.
4. A detector as one signal among others
An AI detector can help decide where to look more closely. The detector on this page estimates how likely the text is to have been generated by AI, marks fragments as likely AI, suspicious or human with a confidence value, gives an overall verdict and produces a PDF report. But it can be wrong in both directions: formulaic, short, heavily revised or second-language texts sometimes get high estimates, and generated text that was later edited can pass as human. That is why the result should never be the only basis for a decision.
5. A fair procedure
- Rules known in advance. Students need to know what is allowed before they start the assignment.
- Evidence weighed together. A change in style, sources that do not exist, a detector result and missing drafts are considered as a whole, never in isolation.
- A chance to explain. The student can show their process and explain the text before any decision is made.
- A proportionate response. Many situations are resolved with a rewrite, an oral presentation or guidance on appropriate AI use.
- A written record. Noting what was reviewed and what was decided protects both the student and the teacher.
For students: how to avoid misunderstandings
Read the course rules, declare any permitted AI use, keep drafts and version history, cite every source you consulted and write in your own voice, with your own examples and opinions. If you are unsure what is allowed, ask before you submit. And if your work is questioned, stay calm: showing your process is the most effective way to clear things up.