Writing with AI at university: a workflow that keeps the work yours
AI can help you write a better essay or get you in trouble, depending on where in the process you use it. A step-by-step workflow for using AI for research, outlining, drafting and editing without losing your voice — or your academic standing.
4 min read
By Kinzo AI Team
- students
- ai-writing
- academic-integrity
- workflow
Most universities now allow some use of AI in coursework and forbid others, and the line is drawn differently in every department. What almost every policy agrees on is the principle: the thinking has to be yours, you have to be able to defend every sentence, and you have to be honest about what help you had.
That principle is more useful than any list of allowed tools, because it tells you where in the writing process AI helps and where it quietly replaces you. Here is a workflow built around it.
Before anything: read the policy
Your module handbook or syllabus almost certainly has a section on AI. Read it. Some courses permit AI for brainstorming and grammar but not for drafting; some require a disclosure statement; a few ban it outright for specific assessments. Five minutes here saves you from an integrity meeting later. If the policy is unclear, email the lecturer and keep the reply.
Stage 1: understanding the question
This is where AI is most useful and least risky. Paste the essay question and ask it to explain what the question is really asking, what a strong answer would need to cover, and what a weak answer would miss. Ask it to list the main scholarly debates on the topic.
Treat the output as a map, not as facts. Everything it says about the literature needs checking. But as a way to get oriented quickly, it is excellent.
Stage 2: research
Do not ask an AI for sources and put them in your bibliography. It will invent references that look real — we have written a five-minute routine for spotting them. Use the library, Google Scholar and your reading list for the actual sources.
Where AI helps is in processing what you find: summarising a dense paper so you know whether it is worth reading in full, explaining a method you do not understand, or pulling out the main claim of a chapter. Read the original before you cite it. A summary is a reason to open the paper, not a substitute.
Stage 3: outlining
Write your own thesis first. One or two sentences, in your words, stating what you will argue. Then use AI to stress-test it: ask what the strongest objection is, what evidence would be needed, and what a sceptical marker would push back on.
You can also ask it to suggest a structure and then rearrange, cut and add until the outline reflects your argument rather than a generic essay shape. If you cannot explain why each section is there, the outline is not yours yet.
Stage 4: drafting
This is the stage where policies differ most and where the risk is highest. If your course allows AI-assisted drafting, use it for the parts that are mechanical — a first attempt at a paragraph you have already outlined and sourced — and write the parts that carry the argument yourself: the introduction, the thesis, the analysis of evidence, the conclusion.
If you do use AI for a first draft of a paragraph, rewrite it. Not because rewriting hides anything, but because a paragraph you have not rewritten is a paragraph you cannot defend in a viva or a follow-up question. Machine-drafted prose also has habits — uniform sentence length, stacked transitions, hedges on everything — that read as exactly what they are. Our list of twelve AI writing tells is a checklist for this pass.
Stage 5: editing
Editing is where AI earns its keep with almost no integrity risk, because the ideas and structure are already yours. Use it to:
- Flag sentences that are unclear or too long.
- Suggest a plainer word where you have reached for jargon.
- Check that each paragraph has one point and that the point is stated.
- Catch grammar and punctuation.
Tools like Kinzo's Humanizer sit here too. Run a paragraph that reads stiffly through it, look at every highlighted change, and keep the ones that improve the sentence. Reject the rest. You are editing, not outsourcing.
Stage 6: checking
Before you submit, two checks.
First, read the whole essay aloud. Anything you stumble over, a marker will too. Anything you do not fully understand — a claim, a term, a citation — has to go or be understood.
Second, if you have used AI anywhere in the draft, run it through an AI detector and read the highlighted sentences. Not because the score decides anything, and not to game it, but because the highlighted sentences are usually the ones that still sound like a template. Fix those and you have a better essay regardless of who reads it. We have written about what detector scores actually measure and why they should never be treated as proof.
Stage 7: disclosing
If your course asks for an AI statement, write a specific one. "I used ChatGPT to clarify the essay question and to check grammar in the final draft; all sources were located and read by me" is a good statement. "AI tools were used" is not. If your course does not ask, keep your own note of what you used and where. If a question is ever raised, that note is your answer.
What this workflow protects
Used this way, AI makes you a faster researcher and a sharper editor without touching the parts of the essay that are being assessed: your understanding, your argument and your judgment. That is also the version of AI use that survives any policy, any detector and any conversation with a marker — because at every stage, you can explain exactly what you did and why.
Kinzo AI Team
The people building Kinzo
We build the Humanizer, Paraphraser, AI Detector and Citation Generator at Kinzo AI, and we write about the writing problems we see every day: fake citations, misread detector scores, drafts that sound like a machine, and how to fix them.