Academic Integrity: Know What Is Allowed and Explain Your AI Use

Permission, disclosure, citation, and authorship answer different questions.

A student adds a sentence to the end of an assignment: “AI helped with this work.” The student assumes that being open about the assistance makes it acceptable.

But the assignment prohibited generated answers. The disclosure describes the assistance without changing the rule.

That fictional situation reveals four questions students need to separate. Was this use permitted? What assistance must be disclosed? Does any material require citation? What work and judgment are being represented as the student’s own?

A clear answer to one does not automatically answer the others. Learning to distinguish them makes it easier to use permitted support, explain your process, and avoid surprises at submission time.

Read the rule at the right level

Your institution may have broad academic-integrity requirements, your course may have an AI policy, and an individual assignment may specify narrower conditions. Read the relevant documents together. If they seem inconsistent, seek clarification rather than choosing the interpretation that permits the most assistance.

Cornell’s teaching guidance encourages instructors to communicate AI rules in syllabi and assignments and clarify documentation and student-authorship expectations. It offers different sample approaches, which illustrates why one course’s permission should not be imported into another. Cornell: AI and Academic Integrity

Look for actions, not just product names. A grammar feature, translator, coding assistant, or search feature may perform different kinds of work. Ask what it does to your submission and whether that operation falls within the rule.

Keep a copy of the instructions you relied on. A short note about an instructor’s clarification can also help, using the institution’s normal communication process.

Separate the four questions

An assignment can permit brainstorming while requiring the student to write every paragraph. Another can permit a generated first draft but assess the student’s evaluation and revision. Both require the student to follow the actual task rather than a personal definition of “help.”

Citation is also different from evidence. Citing a chatbot response does not turn its invented reference into a real paper. Trace factual claims to appropriate sources and check them independently.

Work through six policy scenarios

Use this original fictional policy:

For the local-history essay, students may use the approved assistant to brainstorm questions, explain assigned vocabulary, and flag unclear sentences in a student-written draft. The assistant may not supply submission paragraphs, translate submission text, or generate the bibliography. Students must verify sources themselves and include a short assistance statement. Coding assistance is not addressed. No private interview recording may be uploaded.

Classify each action before reading the decision:

The final answer is deliberately not a yes or no. A policy can leave a real question open. The appropriate next step is a specific inquiry, not an invented permission.

The recording restriction applies independently. Even a permitted brainstorming task does not authorize uploading a private interview.

Ask a question someone can answer

A useful message might say:

The essay policy permits brainstorming and flagging unclear sentences but does not mention coding. May I use the approved assistant to help create a local script that sorts my own source notes by date? It would not draft essay text or create citations. I would review the output and disclose the coding assistance.

This identifies the proposed action, its inputs, and its relation to the assessed work. It also gives the instructor room to set conditions or decline.

An assistant can help compare your proposal with supplied rules:

Classify each proposed use against this assignment policy. Quote only the short relevant clause when necessary, explain the match, and mark ambiguity. Do not treat silence as permission. Help me draft one focused question about the unresolved use.

Review its interpretation yourself. The assistant is not the institution’s decision-maker.

Keep evidence of the work you actually did

Save useful stages: your notes, outline, first draft, selected feedback, and revised draft. Keep references connected to the passages they support. You do not need to preserve every keystroke to show a meaningful process.

If the assistant identified a gap in your argument, record that contribution. If you rejected its suggestion because the source did not support it, record the reason. Both are part of an honest account of the work.

Do not manufacture a draft history after the fact to suggest that assistance never occurred. If your records are incomplete, describe what you know and what you did not retain.

Your ability to explain the work matters too. Can you locate a quotation, justify a method, or discuss why you revised a paragraph? Those are useful demonstrations of understanding, even though they do not by themselves settle every policy question.

Write a specific assistance statement

For a permitted use in the fictional course, an illustrative statement is:

I used the course-approved assistant to generate possible research questions and flag unclear sentences in my first draft. I selected the research question, read and verified the sources, and wrote and revised the essay. I checked each flagged sentence against my intended meaning. The assistant did not supply submission paragraphs or bibliography entries.

Use this wording only if it accurately describes what happened. Replace general labels with the actual tool or required details when the assignment asks for them. Do not say “grammar only” if the assistant proposed a thesis or reorganized the argument.

If a task requires a particular citation style for generated material, follow its current instructions and the relevant style authority. A disclosure paragraph and a formal citation may both be required, because they serve different purposes.

Respond to a concern with the process record

If an instructor questions the work, ask what concern needs clarification and use the institution’s process. Bring the actual source notes and draft history you have. Explain the assistance and your own decisions plainly.

Automated AI-detection output should not be treated as a complete account of authorship. USC Libraries cautions that detection tools can be unreliable and can falsely flag student writing. Cornell’s guidance emphasizes objective evidence and inquiry when investigating concerns. These are reasons to examine the work and process, not reasons to promise that any particular detector result proves innocence or misconduct. USC Libraries: Limitations and Warnings, Cornell: AI and Academic Integrity

Do not rewrite merely to “beat” a detector. Improve the actual work and describe the process accurately. If you used prohibited assistance, disclosure is still important, but ask the instructor how to address the situation rather than relabeling it as permitted.

Carry the habit into professional work

A workplace, publication, grant program, or certification may also define acceptable tools, confidential inputs, and attribution expectations. The same four questions remain useful, although the applicable rules differ.

A student who learns to explain a contribution precisely is practicing a professional skill. “I used this tool for this task, checked these details, and made these decisions” is more informative than either claiming complete automation or pretending no tools were involved.

Practice: classify and explain

Classify the six scenarios above. Write an assistance statement for one permitted use and a clarification question for the unresolved one. Identify a case where permission exists but the evidence or disclosure is still inadequate.

Completion check: You distinguish permission, disclosure, citation, and authorship. Your statement matches the actual work. Prohibited assistance is not relabeled as acceptable through acknowledgment, and ambiguous instructions lead to a focused question rather than a guess.