How should scholars use AI without compromising academic integrity?

Use AI to find, compare, and organize — never to author. It can surface forgotten notes, compare how authors disagree, and locate contradictions in your library. The line that must stay visible is the difference between what you read, what a machine suggested, what you thought, and what you wrote.

Assistant, not ghostwriter

Legitimate uses: surfacing connections across your own material, rediscovering a note from years ago, comparing theoretical positions, clustering themes, finding contradictions you missed, and helping translate finished scholarship for a public audience.

Illegitimate uses: generating claims you have not verified, producing citations you have not read, and drafting argument you then adopt as your own thinking.

The fabrication problem

Language models produce fluent, plausible citations that do not exist. In a scholarly context this is not a quirk — it is a career risk, and it has already ended some.

The defense is structural: never accept a citation that has not resolved against an authoritative source, and prefer 'unable to verify' over a confident guess.

Keep the record

Every AI-touched item should carry a marker: what model, when, from what prompt, and whether a human verified it.

That record turns an uncomfortable question — 'did you write this?' — into one you can answer with evidence.

Common questions

Can AI write my literature review?

It should not. It can help you organize and compare sources you have actually read; the synthesis and the argument must be yours.

Why do AI tools invent citations?

They generate statistically plausible text, and a plausible-looking citation is easy to produce. Always verify against an authoritative index before citing.