Bans produce a compliance-and-detection dynamic. Students who use AI conceal it, faculty who suspect it run detectors, and the classroom becomes an adversarial workflow around a tool that is easier to hide than to detect. Disclosure changes the incentive. If saying you used AI for brainstorming carries no penalty, students say it. If saying you used AI to generate the final paragraph carries a defined and moderate consequence, students learn where the line is.
Quick answer
Introduce the disclosure norm at term start, provide a simple template students can copy, distinguish acceptable use (brainstorming, revising, citation help) from unacceptable use (generating final text), grade a short reflection alongside each major assignment, and handle resistance by making disclosure easier than concealment. The whole system takes about thirty minutes of class time to set up.
The rest of this piece walks through why the norm works where bans do not, how to introduce it in the first week, a template students can adapt, the acceptable/unacceptable line, and the resistance patterns to expect.
Why disclosure works where bans don't
The larger effect is on the middle ninety percent. Detection-first policies treat every student as a potential cheater. Disclosure-first policies treat every student as capable of reporting on their own process. The second frame produces better writing about writing, which is itself a useful academic skill.
Disclosure treats every student as capable of reporting on their own process. Detection treats every student as a suspect. That difference sets the tone of a course.
The Working Educators editors
How to introduce the norm
The disclosure norm needs to appear in the syllabus, be discussed on the first day of class, and be reinforced on the first assignment. All three matter. A syllabus line alone is invisible. A first-day discussion without reinforcement fades.
The first-day discussion should take ten to fifteen minutes. Explain what counts as acceptable use in your course, walk through an example disclosure statement, and answer the two questions students always ask: what if I used AI and forgot, and what if I am not sure whether something counts. Reinforce that the disclosure itself is the safe path.
A simple disclosure template
At the bottom of each assignment, students submit a short paragraph in this form: I used [tool name] for [specific task], with [specific prompt or approach]. The output influenced [specific aspect of the final work]. I did not use AI for [specific unacceptable uses, per the course policy].
One paragraph, three sentences. Students who used AI for brainstorming say so and get credit for transparency. Students who did not use AI say so explicitly, which is also useful information. The template deliberately does not require students to attach prompts or transcripts. That would produce compliance theater rather than honest reporting.
What counts as disclosed vs undisclosed
The line varies by course. For a first-year writing seminar, acceptable might include brainstorming, outlining, and grammar checks, with generation of thesis statements and body paragraphs off-limits. For an upper-division elective, acceptable might extend to first-draft summaries with substantial rewriting by the student. For a lab report, acceptable might be almost anything on the discussion section as long as the data reporting is untouched.
The specific line matters less than the fact that a line exists, is written down, and is discussed. Our classroom AI policy guide walks through the syllabus language that supports this.
Handling the common resistance
The most common resistance is not defiance. It is forgetting. Students genuinely do not remember what they used AI for a week ago, especially in a heavy load of courses. The fix is to require the disclosure at the same moment the assignment is submitted, not as a separate follow-up.
The second pattern is over-disclosure. Some students, especially early in the term, disclose everything including uses that were plainly incidental, like a grammar check on a single sentence. Reassure them that the goal is honest reporting, not exhaustive logging. The third pattern is under-disclosure by students who do not think a use counts. This is where the syllabus examples do the work. Make the examples specific enough that students can pattern-match to their own behavior.
For the small number of cases where a student clearly used AI in a prohibited way and did not disclose it, the response should be proportional and documented. The disclosure norm gives you a better record for those cases than a pure detector-based system, because the absence of disclosure combined with corroborating evidence is a much stronger signal than a detector score alone. See our guide for handling denials for the conversational framework.
Frequently Asked Questions
Does disclosure replace AI detection entirely?▼
In most course contexts, yes. Detection can play a small triage role, but the primary integrity signal comes from the presence or absence of an honest disclosure paragraph.
What if students lie on the disclosure?▼
Some will. The norm still improves outcomes because a false disclosure combined with corroborating evidence, like a rubric that rewards course-specific engagement, is a stronger integrity case than a detector score alone.
How long does the disclosure paragraph take students?▼
Two to three minutes per assignment after the first one. Students internalize the format quickly and write faster over the term.
Should the disclosure be graded?▼
The presence of a disclosure should be a small grade item, effectively a completion check. The content of the disclosure is not typically graded except as it relates to the assignment's other criteria.
What if my institution requires a stricter policy?▼
Disclosure is compatible with almost any institutional policy, including strict ones. It surfaces the information the institution needs to enforce the policy. If your institution mandates a specific detector, run disclosure alongside it and prioritize the disclosure record in adjudication.
The Bottom Line
Disclosure is the cheapest, least adversarial integrity tool available in the AI era. It costs thirty minutes of setup time and produces a durable norm that survives model updates in a way that detector-based policies do not.
The version of this that has held up in classrooms combines a written syllabus line, a first-day discussion, a per-assignment template, and a rubric that rewards course-specific engagement. None of those requires a detector. All of them require the instructor to make the norm visible and easy.
