UVM AI Detection Accusation: What the Code of Student Conduct Requires

A University of Vermont AI detection accusation triggers the Center for Student Conduct process under the Code of Student Rights and Responsibilities. The detector score is a flag, not a finding. What resolves these cases is process evidence: drafts, version history, handwritten notes, and the specific policy language your institution actually applies.

The pattern at the University of Vermont

UVM instructors have access to Turnitin through the Brightspace learning environment, and Turnitin's AI writing indicator is part of the standard similarity report. The pattern in AI-detection cases at UVM tracks what has been reported at peer institutions: an instructor sees a Turnitin AI percentage, forwards the report to the student with a request to meet, and either handles the matter at the instructor level or refers it to the Center for Student Conduct.

History, English, political science, and other essay-heavy humanities courses generate the majority of these referrals. Lab reports and formal technical writing also trigger flags. The reason is not that these students used AI; it is that the writing genres share statistical features with model output. Our breakdown of what AI detectors actually measure covers the mechanics.

Why detectors flag human writing

Turnitin and similar tools read text through two statistical lenses: perplexity (how predictable each word choice is) and burstiness (how much sentence length and complexity vary). Careful academic prose written by a student who has revised for clarity often lands in the low-perplexity, low-burstiness zone that the detector was trained to associate with generated text.

Peer-reviewed research has documented the consequences. Weber-Wulff and colleagues (2023), publishing in the International Journal of Educational Integrity, tested fourteen detection tools and concluded none met the reliability threshold expected of evidence in an academic proceeding. Liang and colleagues at Stanford (2023), publishing in Patterns, found that AI detectors misclassified TOEFL essays written by non-native English speakers at strikingly high rates while flagging native-speaker essays much less often. If English is not your first language, that finding is directly relevant.

What UVM's Code of Student Rights and Responsibilities requires

The University of Vermont handles academic integrity through the Code of Student Rights and Responsibilities, administered by the Center for Student Conduct. The Code prohibits academic dishonesty, which includes submitting work generated by another source (including generative AI where prohibited by the instructor) and representing it as your own. The relevant questions in an AI case are almost always narrower than the accusation makes them sound:

• Did the specific course syllabus or assignment prompt prohibit the use of generative AI?
• What evidence supports the claim that generative AI was used, beyond the detector score?
• Was the standard of evidence required by the Code actually met?
• Did the student have an opportunity to respond before a finding was issued?

UVM's conduct process gives you the right to review the evidence against you, to respond in writing or at a meeting, and to appeal a finding on defined grounds. Read the syllabus for the course carefully. AI policies at UVM are set at the instructor level, and what is prohibited in one course may be permitted or required in another.

Evidence that shifts these cases

Process evidence closes AI-detection cases more often than argument does. The strongest artifacts are the ones a generative tool cannot produce after the fact:

• Google Docs or Microsoft Word version history showing the paper built up over time, with pauses, deletions, and revisions consistent with human drafting
• Handwritten notes, outlines, and annotated readings photographed with visible dates
• Browser history for the drafting window showing library database access, source lookups, and citation formatting
• Bates, Howe, or MultiPass library records and citation manager entries (Zotero, EndNote) matching the sources cited in the paper
• Messages with classmates, TAs, or the writing center discussing the assignment during the drafting period

Do not edit the flagged document. Every save can overwrite version history and change metadata. Our step-by-step evidence gathering guide covers what to preserve in the first 48 hours, and the Google Docs version history walkthrough shows how to export it in a form the Center for Student Conduct will accept.

If this is you at the University of Vermont

Move deliberately, in this order:

1. Ask, in writing, for the specific detector used, the exact score reported, and any notes from human review. Save the response.
2. Preserve every draft, version history export, note, and browser history entry before default retention windows close.
3. Read the course syllabus and the Code of Student Rights and Responsibilities. Confirm what was actually prohibited and what standard of evidence applies.
4. Do not admit to something you did not do to make the meeting easier. An informal resolution is still a finding on your record.
5. If the proposed sanction is suspension or dismissal, or if you are an international student whose visa status depends on continued enrollment, consult an education law attorney before your meeting. Your procedural rights FAQ covers what you can request before that meeting.

If you are preparing a written response for a UVM Center for Student Conduct meeting, NotBot generates a personalized defense package that names the detector that flagged you, cites the research on its accuracy, and organizes your process evidence into a form conduct officers can read quickly. If you have already had a finding issued and are past the initial response stage, the appeal package is built around the grounds UVM's appeal process actually accepts.