Suspected AI-Written Work: A Fair Referral Process From Classroom to Office
Published on October 6th, 2026 by the GraideMind team
Few moments in a teacher's week feel as uncomfortable as suspecting that a student's essay was not written by the student. The writing may suddenly be polished, the voice may have changed, or a quotation may not appear in the text. Teachers worry about accusing an honest student, and also about letting dishonesty pass. Institutions that offer a clear process make these moments far less fraught.

The most important principle is that suspicion is not proof. Detection tools are known to produce false positives, and multilingual writers can be flagged at higher rates, so a score alone should never be the basis for a penalty. Evidence of the writing process, a conversation with the student, and consistency with earlier work are far stronger. Treating the initial concern as a reason to look further, rather than a verdict, protects everyone.
Start with the assignment design. Policies that explain what tools are permitted for each task, and that include process evidence such as outlines and drafts, make it easier to resolve questions fairly. Students who know the rules and how authorship will be verified are less likely to cross the line. The best referral process is one that is rarely needed.
Begin with a conversation, not an accusation
The first step is usually a private, curious conversation. Ask the student to describe how they developed their argument, to explain a particular passage, or to walk you through their draft history. A student who wrote the paper can usually discuss it comfortably, even if they are nervous. Keep notes of what was said and be open to the possibility that there is an innocent explanation.
- State plainly what AI use is permitted on each assignment
- Treat suspicion as a reason to ask questions, not as proof
- Never rely on a detection score alone when deciding on a referral
- Collect version history, notes, and earlier writing samples before escalating
- Keep referral summaries factual and separate from the scoring of the work
A fair integrity process starts with curiosity and ends with evidence that more than one person could examine.
Stop spending your evenings grading essays
Let AI generate rubric-based feedback instantly, so you can focus on teaching instead.
Try it free in secondsGather evidence before escalating
Collect the documents that bear on the question, such as version history, notes, previous writing samples, and the assignment instructions. Compare the work in question with earlier submissions, bearing in mind that students improve and that one strong essay is not suspicious by itself. Be cautious about relying on any single indicator. A fair case rests on multiple, consistent pieces of evidence.
Know your institution's procedure before you need it. Many schools require teachers to document concerns and refer cases to an administrator or integrity officer, and some track referrals to look for patterns. Following the process protects the student's rights and your own. If the policy is unclear, ask your department head or an administrator to clarify, and learning the steps in advance means you will not be improvising a response while a student is waiting for an answer.
Document clearly and keep records factual
When a referral is warranted, submit a concise, factual summary. Describe what you observed, the evidence you gathered, what you asked the student, and how they responded. Avoid language that speculates about motives or character. Facts allow administrators to make fair decisions, and they help ensure consistency across cases, and a reviewer who reads a calm, organized summary can reach a fair conclusion far more quickly than one sorting through impressions.
Attach the assignment, the rubric, and any relevant policies to the referral so the reviewer has context. If a rubric-based tool was used for grading, note that scoring is separate from the integrity question, and that the tool does not determine authorship. Scores should not be adjusted as punishment outside the established process. Keeping grading and discipline distinct is a mark of fairness.
Support students through the outcome
Whatever the result, treat the student with respect. If the concern is substantiated, a restorative approach that includes a chance to redo the work and learn from the situation often serves the student better than punishment alone. If the concern is unfounded, acknowledge it and repair the relationship. Students remember how they were treated at moments like this, and a follow-up check-in a few weeks later tells the student that the relationship has not been defined by a single incident.
Use what you learn to improve your assignments. Cases often reveal that a prompt was easy to automate or that expectations were unclear. Adjusting the task by adding in-class components, personalizing topics, or requiring process evidence can reduce future problems. The process becomes a source of improvement instead of just a source of stress, and sharing the changes with colleagues spreads the benefit beyond a single classroom to the whole department.
See how fast your grading workflow can be
Most teachers go from hours per batch to minutes.
Create free account


