Drafting an essay with ChatGPT and keeping it yours

1Uses most policies allow

University policies vary, but many follow a pattern like UCL's. In its assistive category, students may use generative AI in a limited, critical and responsible way, and the student should still be the author. UCL's examples include supporting the writing process in a limited manner, such as proofreading or reviewing academic tone, acting as a support tutor, and giving feedback.

In practice that usually covers asking a tool to explain a concept, suggest angles you could research, question your outline, or point out unclear sentences in a draft you wrote. The Russell Group's principles, agreed by its 24 universities in 2023, also commit them to helping students use these tools appropriately and to letting students ask questions about specific uses without fear of penalty.

2Uses that cross the line

• Pasting generated paragraphs into your essay as your own writing.
• Asking the tool to write the essay and then editing it lightly.
• Using it to rewrite a source so it no longer looks copied.
• Citing sources it suggested that you have not found and read.
• Any use at all in an assessment where AI is not allowed.

UCL notes there are no circumstances where students may use generative AI to gain unfair academic advantage, and treats misuse as plagiarism or falsification. For the plagiarism question in detail, see is using AI plagiarism?

3A workflow that keeps the essay yours

1. Read the brief and the AI rule first. If AI is not allowed, stop here.
2. Write your own thesis before opening a tool. See thesis statements for why tool-written ones sound generic.
3. Plan with questions. Ask the tool what objections your thesis faces, or what you might be missing. Decide yourself what to use.
4. Draft alone. Write every paragraph yourself from your own notes and sources, following essay structure.
5. Ask for feedback, not rewrites. See AI feedback on drafts.
6. Check every fact and source it mentioned against the original.

4One essay, two workflows

Imagine an essay asking whether city centres should ban cars. In the first workflow, the student asks a chatbot to write an essay on the question, edits the wording and submits it. The argument, the structure and most of the sentences came from the tool. In most courses that is misconduct, and the essay is usually generic too.

In the second, the student reads three sources, writes a thesis (a ban helps air quality but hurts small shops unless deliveries are allowed), and drafts the essay alone. Where the course allows assistive use, they ask a tool for the strongest objection to the thesis and for sentences that are hard to follow, then decide what to change and change it themselves.

The second essay took longer, but every decision in it belongs to the student, and they can explain all of it if asked. That is the difference most policies are trying to protect.

5What to keep as a record

Keep your notes, outline, drafts and version history, and a short log of any tool use: date, what you asked, what you kept. If your course asks for an acknowledgement, you will have what you need to write it. If your work is ever questioned, a visible drafting history is far more persuasive than anything you can say afterwards.

Admissions essays follow different rules again: see admission essays and AI.

More on this in AI and writing.

6FAQ