We sent public records requests to 200 universities in the United States and United Kingdom. We asked a simple question: if a student is accused of submitting AI-generated work based on a detection tool's output, what is the formal process for appealing that accusation? The answers - and the silences - paint a picture of an institutional crisis that most universities have not yet acknowledged.
Of the 200 institutions contacted, 163 responded with usable information. Of those 163, only 41 have a published appeal process that specifically addresses AI detection disputes. The remaining 122 either direct students to generic academic misconduct procedures or have no formal process at all.
The Three Categories
Category 1: Dedicated AI appeals (41 institutions). These universities have created specific procedures for disputes arising from AI detection flags. The best of them include: a right to see the full detection report, a guaranteed timeline for resolution (typically two to four weeks), the right to present evidence of authorship, review by someone other than the original instructor, and a presumption of innocence until evidence is evaluated.
Category 2: Generic misconduct procedures (87 institutions). These universities route AI detection disputes through existing academic integrity frameworks that were designed for traditional plagiarism. The problem is structural: plagiarism procedures assume that evidence of copying exists (matching text, uncited sources), while AI detection disputes hinge on statistical probability. A process designed to evaluate whether a student copied from a source is poorly equipped to evaluate whether a statistical model's confidence score is reliable.
Category 3: No formal process (35 institutions). These universities mandate AI detection but provide no published mechanism for contesting a flag. In practice, disputes are resolved informally - between the student and the instructor - with no guaranteed timeline, no independent review, and no documentation standards. The outcome depends entirely on the instructor's willingness to listen.
The Timeline Problem
Among institutions with formal processes, resolution timelines vary dramatically. The fastest (a small liberal arts college with a dedicated academic integrity office) averages eight days. The slowest (a large research university routing disputes through a faculty committee that meets monthly) averages eleven weeks. For students, those weeks are not abstract: they represent missed deadlines, deferred grades, suspended financial aid, and the psychological weight of living under accusation.
Several students we interviewed described the waiting period as worse than the accusation itself. "I could handle being told I was wrong," one undergraduate at a midwestern state university told us. "What I couldn't handle was six weeks of not knowing whether my academic career was over."
The Evidence Problem
Even when appeal processes exist, they often fail at the most basic level: defining what constitutes sufficient evidence of human authorship. We found no institution that publishes clear evidentiary standards for AI detection disputes. Students are told to "present evidence" but are not told what qualifies. Version history? Research notes? Testimony from an advisor? Screen recordings of the writing process? Each case is evaluated ad hoc, by reviewers with varying technical literacy and varying familiarity with how AI detection tools work.
This ambiguity creates two problems. First, students who are accused do not know what to prepare. Second, reviewers do not know what to evaluate. The result is inconsistency: identical cases resolved differently depending on who reviews them.
Recommendations
Based on our analysis, we propose a minimum standard for AI detection appeal processes. Every institution that uses AI detection should publish: a specific appeal procedure separate from general misconduct processes; a guaranteed resolution timeline of no more than four weeks; clear evidentiary standards defining what constitutes proof of human authorship; mandatory independent review by someone other than the accusing instructor; a right to access the full detection report including the tool used, score, and flagged passages; and annual reporting on the number of flags, appeals, and outcomes, disaggregated by student demographics.
These are minimum standards. The best institutions will exceed them. But at present, the majority of universities mandating AI detection have not met even the baseline. Until they do, the system remains one in which the accused bear the burden, the process is opaque, and the outcomes are arbitrary. That is not integrity. It is institutional failure.