Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Start with the rules that applied to the assignment. Using AI is not automatically misconduct: the key question is whether the student’s use exceeded what the course allowed, omitted a required disclosure, or replaced work the assessment was meant to measure. A writing style or AI-detector score cannot establish that on its own. Check the evidence, talk with the student, document the exchange, and follow your school’s academic-integrity process.
What counts as inappropriate AI use?
There is no single rule that applies to every class or school. The syllabus, assignment instructions, institutional policy, and any relevant jurisdictional rules determine whether a particular use was permitted and whether it had to be disclosed. For example, the University of Toronto advises instructors to state whether and how AI can be used; UMass Amherst says students need instructor permission under its policy. Those are local examples, not universal rules.
Compare the student’s use with the expectations in force for that assignment. Consider whether the assignment was intended to assess the student’s own reasoning, research, drafting, or subject knowledge, and whether the student followed any disclosure or citation requirement. Do not apply a new or unclear restriction retroactively.
For guidance on setting expectations, see the University of Toronto’s generative AI teaching guidance and UMass Amherst’s AI and academic integrity guidance. Rules can also be jurisdiction-specific: NSW HSC requirements, for example, apply within that assessment system.
Which signs justify a closer look?
Look for observable, checkable discrepancies—not a label such as “AI-written.” Official guidance identifies possible leads such as:
- References that cannot be verified or do not support the claims attributed to them.
- Factual errors, irrelevant material, or an answer that does not address the assignment.
- Text that does not engage with course content or differs markedly from the student’s demonstrated understanding.
- A prompt repeated in the answer, or unusual revision or submission timing.
These clues do not prove misconduct. The University of Rochester explicitly cautions that none of its listed indicators is conclusive by itself. An apparent change in vocabulary or fluency can have explanations unrelated to AI use, so assess it alongside the assignment requirements, sources, course work, and the student’s account.
Check for information that cuts against your initial concern as well as information that supports it. The Australian higher-education regulator TEQSA warns that confirmation bias can distort an investigation; a fair review actively looks for disconfirming evidence.
Why an AI detector cannot decide the case
Detector output is not a probability that a particular student cheated. Tools can produce false positives and false negatives, disagree with one another, or perform less reliably on short, edited, or mixed human-and-AI text. Some guidance also raises concerns about bias and privacy.
Institutional approaches differ. The University of Toronto says it does not support AI-detection software for student work, citing reliability, privacy, and ethical concerns. Rochester does not recommend detectors, and UMass Amherst says its Academic Integrity Office does not recommend relying on detector reports. TEQSA and NSW’s education authority, NESA, describe more limited or cautious roles for such tools under local conditions. The University at Buffalo mentions Turnitin among possible investigative tools but says a report alone is insufficient. Follow your own institution’s policy; do not use a score as the sole basis for a misconduct finding.
TEQSA offers a hypothetical illustration, not a measured result for a current detector: if a detector had a 1% false-positive rate, it would flag one assignment in 100 as having a high score, such as 80–90%. That example illustrates why a high score does not mean there is an 80–90% chance that a student used AI.
How to handle a concern fairly
- Check the applicable rules. Read the syllabus, assignment prompt, institutional policy, and any applicable jurisdictional requirements. Note what AI use was allowed, what had to be disclosed, and what the assessment was designed to measure.
- Write down the specific concern. Identify the exact passage, source, claim, or other observable feature that needs explanation. Verify references and compare the submission with relevant course material or prior work where appropriate. Avoid treating style or a detector score as proof.
- Invite the student to discuss the work. Explain what you want to understand and ask the student to walk through their sources, reasoning, and writing process. Keep the conversation open enough for them to explain the work or correct a misunderstanding.
- Record both the concern and the explanation. Make a factual note of the applicable assignment rule, the material that raised concern, checks or comparisons you made, the student’s response, and the next step.
- Use the school’s process if concerns remain. Consult or report to the designated academic-integrity office under local rules. Do not announce a finding or impose a sanction outside the procedure your school requires.
Questions that keep the conversation focused
- “Can you explain what you mean by this term?”
- “Walk me through how you got to the conclusion of your paper.”
- “How did you go about finding these sources?”
- “What was your writing process like?”
A student’s difficulty answering one question is not a finding. Consider the context of the conversation, including anxiety, disability or language needs, and the institution’s process. Ask about the work rather than staging an interrogation around a detector result.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What to do when evidence is unclear
If the rule was ambiguous, evidence is inconclusive, or the student’s explanation resolves the concern, do not turn suspicion into a disciplinary finding. Consult the appropriate office if you are unsure how the local procedure applies, and clarify expectations for future assignments. If concern remains, refer it through the established process so the student receives the notice and review required by that institution.
Resolution routes and evidentiary thresholds vary. The University of Rochester and University at Buffalo describe procedures for their own settings; those procedures should not be assumed to apply at another school.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to reduce ambiguity in future assignments
State permitted and prohibited uses in both course and assignment instructions. Specify whether the rules cover generative chatbots, writing editors, summarizers, or other tools, and explain whether students must acknowledge or cite permitted use. Connect the rules to the learning outcome: students are more likely to understand a boundary when they know what the assessment is measuring.
Where it fits the learning goal, use more than one way for students to demonstrate understanding—for example, staged drafts, short in-class writing, an oral explanation, or follow-up questions about submitted reasoning. These are assessment methods, not traps. Toronto recommends asking students to expand on out-of-class work; Rochester discusses oral discussion and short in-class writing; NESA calls for varied assessment tasks.
TEQSA discusses a “two-lane” assessment approach: some key assessments can be designed to verify learning outcomes under more secure conditions, while other learning-focused tasks may allow AI use with acknowledgment. It is one design framework, not a universal requirement. Choose assessment formats that fit your course outcomes and local policy.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
