AI and dissertation integrity: preserve your authorship

§ 1 · Start here

Treat local policy as the controlling instruction

AI rules vary across institutions, programs, committees, courses, publishers, data agreements, and research settings. A practice permitted in one context may be restricted or require disclosure in another.

Before using an AI tool, read the current policies that apply to your work and ask your chair when the language is unclear. Follow the most specific applicable instruction. Do not assume that a tool’s availability, a colleague’s practice, or a general online guide grants permission.

• Institutional academic integrity and acceptable-use policies
• Doctoral program, dissertation handbook, and committee expectations
• Course or milestone instructions that govern the current work
• IRB, participant-consent, sponsor, data-use, or confidentiality terms
• Journal, conference, or repository disclosure requirements

§ 2 · Core boundary

Preserve authorship at the level of reasoning and expression

Authorship is more than typing words. It includes deciding what matters, selecting and interpreting evidence, making methodological choices, forming claims, acknowledging limitations, and taking responsibility for the final text.

Use AI, when permitted, to support reflection rather than to replace those decisions. Questions, counterarguments, checklists, and planning prompts can help you inspect your thinking. Ready-to-submit passages, invented citations, undisclosed paraphrases, or interpretations you cannot defend weaken your authorship.

• Can I explain and defend every claim without relying on the tool?
• Did I read the cited source rather than trust an AI summary of it?
• Are the interpretation and limitations genuinely my judgments?
• Is the final wording my own scholarly expression?
• Could I accurately describe this use to my chair?

§ 3 · Practical workflow

Use a permission–purpose–proof check

Before each use, write down three things: permission, purpose, and proof. Permission identifies the rule or approval that allows the use. Purpose states the narrow task. Proof records what you checked, changed, rejected, or learned.

This short record helps prevent tool use from quietly expanding. It can also support an accurate disclosure if your program requires one. Documentation does not make a prohibited use acceptable; it supports transparency for uses that are permitted.

§ 4 · Safety check

Verify outputs and protect sensitive material

AI output can be incomplete, inaccurate, or fabricated. Verify factual claims against primary or authoritative sources, open and read every citation, check calculations independently, and reject language you cannot substantiate.

Do not enter participant data, identifiable records, unpublished committee material, proprietary datasets, or other protected content into a tool unless the applicable approvals, agreements, and tool configuration explicitly permit it. When uncertain, do not upload the material; ask the responsible institutional contact.

• Verify quotations, authors, titles, dates, identifiers, and page numbers.
• Trace methodological advice to sources accepted in your field.
• Check whether prompts or uploads are stored, reviewed, or used for training.
• Remove sensitive details only when de-identification is allowed and sufficient.
• Keep a record of consequential outputs and your verification steps.

§ 5 · Before you proceed

A five-question integrity check

If any answer is no or uncertain, pause. Narrow the use, remove the sensitive material, verify the output, or ask your chair or the appropriate institutional office before continuing.

• Is this use explicitly permitted, or have I clarified the policy?
• Does the task support my thinking rather than replace my authorship?
• Can I protect all confidential, personal, and restricted information?
• Will I verify every claim, source, and calculation I rely on?
• Can I document and disclose the use accurately if required?