Addressing AI Misuse

AI generated content has become increasingly difficult to identify. While there are some tools that might serve useful, clearly communicating expectations and direct conversations with students are more effective for dealing with issues of AI misuse in class than relying on AI detectors.

It is important to emphasize to students that limiting use of AI in courses and on assessments is an issue of allowing students to demonstrate learning and mastery of a subject. Our job as instructors and professors is to ensure that our students are learning and the assessments we use are one way we use to verify this learning. If the student cannot demonstrate the knowledge, skills, or attitudes required of the class, then the grade assigned and the degree granted are of limited use.

Best Practices for Handling Suspected AI Misuse

These best practices have been gathered from guidance found at Harvard, Cornell, NYU, MIT, Johns Hopkins, UC Berkeley, Princeton, Columbia, ASU, USC, and UCSC, as well as from the presentation “Investigating Misuse of Artificial Intelligence” presented during the ICAI Summer Intensive Series on 25 June, 2026 by Drs. Justin Tonra and Michael Fernbacher.

Before There are Issues

Anticipate problems: Set clear expectations about AI use for your class in the syllabus and on specific assignments. Discuss these with the class. Don’t simply state whether AI can or cannot be used, but define what is meant by GenAI in your class. Does it include editors like Grammarly, LLMs like ChatGPT, or note summarizers like NotebookLM? Remove ambiguity about what students can and cannot use and explain how the use of AI affects your learning outcomes for the class. What is obvious to an instructor, may not be to our students.

Attempt to verify learning: Well-designed assignments should allow students to demonstrate their learning to you. Finds ways to confirm that an assignment the student has submitted demonstrates mastery of your desired learning outcomes. Out-of-class assignments are the most susceptible to unauthorized AI use, so find ways to confirm that what the students have submitted is their work. Oral presentations and discussion or short, in-class writing can expose a disconnect between what the student has submitted and what they claim to know in the assessment. Reach out to the Learning Center if you would like to discuss your assessment strategy.

When you Think an AI Issue has Arisen

Inform the Board: As soon as you suspect a submission was AI-generated, contact the Chair of the Board of Academic Honesty at college.honesty@ur.rochester.edu to indicate that you suspect a policy violation in your class. Both the Chair and the Honesty Liaison  (honestyliaison@ur.rochester.edu) can offer assistance at any point in the process.

Prepare for the conversation with the student: If you suspect a submission was AI-generated, identify the problematic aspects of the submission. What about the submission gave you pause? Make note of these things. These notes will help you when you discuss your concerns with the student and when you submit a report to the board. Among potential signs of AI misuse are:

  • Fabricated/unverifiable/hallucinated sources and citations
  • Anomalies in the revision-history or submission timeline (e.g. unnaturally short turn-around from rough draft to final draft)
  • Discrepancies in level of mastery or proficiency across various assessments (between in-class performance, exams, and papers, for example)
  • Repetition of the question or prompt
  • Inconsistent fonts
  • Superficially polished/error-free writing
  • Unexpected/unusual/archaic vocabulary or unusual phrasing of common terms
  • Introduction, summaries, conclusion in random places (second introductions or two conclusions, as if two papers were pasted together)
  • Failure to engage with course sources or engaging with unexpected sources

Note 1: None of these are by itself conclusive, but it is fair to ask about these issues if you see them.

Note 2: LLMs have learned “to write” from human-generated text, and our students, in turn, are learning to write by reading AI-generated text, so some of the rumored tells of AI use, like the use of em-dashes or the word “delve,” don’t actually reveal much.

Note 3: The University of Rochester does not endorse or recommend any AI Detector software. Although there are better and worse AI detectors, their probative value, in general, is quite low. Different detectors may suggest different probabilities. AI-humanizers can outwit the detectors. Additionally, a positive hit (or a “more likely” hit) for AI usage may sway you to look for and identify things as AI use that might not actually be AI use. Students know this and may use the presence of AI detector data to challenge the accusation. Finally, do not submit personally identifiable information to an AI-detector, as this may constitute a FERPA violation. For more information regarding the drawbacks of AI detection, see the article: Heads I Win, Tales You Lose: AI Detectors in Education

Talk with the Student: Meet as soon as possible with the student after you notice something. Students often claim to forget aspects of their writing process or the material if asked even a few weeks after the suspected behavior. Tell the student that there were certain aspects of the assignment you are concerned about and you want to give them a chance to explain these issues. Ask about their writing/homework/study process. Ask about the content of the paper/assignment. Ask about the aspects of the assignment that originally made you suspect AI misuse. Give the student an opportunity to genuinely explain the work before accusing them of any misconduct. Although it can be challenging, it’s important to be as non-adversarial as possible. It could be an honest mistake by the student and this conversation could help them see that. You are still required to report the misconduct, but at least the student will have had the chance to be heard.

Some questions you might use are:

  • “You use this term, can you explain what you mean by it?”
  • “Explain this particular theory that you use in the paper.”
  • “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?”

Report your findings to the board: Depending on what you discover during the discussion, you have the option to file 1) a warning letter, 2) an instructor resolution with penalty, or 3) a board resolution. Consult with the Chair or the Liaison regarding which option is appropriate for your case. Whichever reporting mechanism you use, document what you noticed and what the student said regarding those issues. Document with sufficient clarity and detail so that a board of non-experts can understand what happened. Consult with the Chair of the Board regarding any grade penalty to ensure consistency across cases in the University.