Should You Use AI for This? A Student’s Decision Guide

AI can do a remarkable amount of university work. It can outline an essay, solve a statistics problem, generate code, summarise a reading, rewrite a paragraph, suggest research questions and produce a polished answer in seconds.

That creates an easy trap. Students start by asking whether AI can do the task, then treat capability as permission. The more useful question is whether AI should do this part of this task, given what the assessment is designed to teach, what the rules permit and what you need to remain able to explain yourself.

This decision belongs before the first prompt. Good AI use starts with choosing the right division of labour between you and the tool.

The Quick Answer: Ask Five Questions Before You Use AI

Before you open ChatGPT, Claude, Gemini or another generative AI system for university work, ask five questions:

1. Is AI allowed here?
2. What am I supposed to learn or demonstrate?
3. What thinking would I lose if AI did this?
4. Could I explain and defend the result without AI?
5. What is the smallest useful role AI can play?

Those questions separate four very different situations. Sometimes the right decision is to do the task yourself. Sometimes AI belongs around the task as a tutor, critic or practice partner. Sometimes it can sit inside the workflow. In assessments where generated content is explicitly permitted, it can form part of the submitted work, provided you follow the stated rules on acknowledgement and responsibility.

AI Fluency Starts Before the Prompt

The AI Fluency Framework describes Delegation as thoughtfully deciding what work to do with AI and what to do yourself. That is an important shift because most advice about AI starts too late. It begins with prompts, settings and techniques after the decision to use AI has already been made.

For students, delegation is partly about efficiency, but it is also about learning and intellectual ownership. A tool that saves time can still be a poor choice if the time it saves is the practice that develops the skill being assessed.

This is why the first step in Vappingo’s Student AI Loop is DECIDE. The broader loop is DECIDE, EXPLAIN, TEST, VERIFY, USE. This guide takes that first step and turns it into a practical decision process you can apply to almost any academic task.

The Five-Question AI Use Check

The five questions work best in order. Permission comes first because no personal judgement about usefulness overrides an assessment rule. The next three questions test whether AI would weaken the learning or your ownership of the work. The final question finds the least intrusive role that still gives you useful help.

1. Is AI Allowed Here?

Start with the specific assessment, not a general impression of your university’s attitude towards AI. Rules often vary by module, task and assessment type. A use that is acceptable for revision can be prohibited in a submitted assignment, while another assessment can require students to use AI as part of the task.

Oxford’s current policy on summative assessment requires each specific assignment to state whether and how students can use AI. Students are expected to follow that declaration and, where AI use is authorised, acknowledge it in the format prescribed by the assessment setter. Read Oxford’s policy on AI use in summative assessment.

UCL uses three broad categories: assessments where GenAI cannot be used, assessments where it can play an assistive role, and assessments where it has an integral role. UCL also makes clear that the instructions in the individual assessment brief take precedence over the general categories. See UCL’s GenAI assessment categories.

Policies differ between institutions, so do not copy another student’s rule of thumb. Check the brief, module handbook, course guidance and any AI statement attached to the task. If the instructions are unclear, ask the lecturer or module team before submitting the work.

2. What Are You Supposed to Learn or Demonstrate?

Once you know the rules, identify the purpose of the task. Ask what your lecturer expects you to be able to understand, perform or defend when the assessment is finished.

If a statistics exercise is designed to test whether you can select and apply the correct method, asking AI to choose the method and solve the problem hands away the skill being assessed. If the assessment is about interpreting a dataset and approved tools are allowed for code support, using AI to help debug a syntax error can leave the central analytical judgement with you.

The same distinction appears in writing. If the task assesses your ability to build and defend an argument, asking AI to generate the thesis, structure and supporting points removes a large part of the assessed thinking. Asking it to challenge an argument you have already developed can create useful pressure without taking ownership of the argument away from you.

This question is especially important because AI can produce outputs that look more advanced than the student’s underlying understanding. A polished paragraph is not evidence that the learning happened. The real test is whether you can reproduce the reasoning, apply it in a different context and respond when someone challenges it.

3. What Thinking Would You Lose If AI Did This?

University learning depends on activities that can feel inefficient: trying, retrieving, comparing, making mistakes, revising, testing an idea and discovering that your first approach was weak. AI can remove some of that friction. Sometimes that is helpful. Sometimes the friction is where the learning is taking place.

Ask what practice disappears when AI takes over the task. If AI turns your lecture notes into quiz questions, you still have to retrieve the answers. If it gives you the answers before you try, the retrieval practice disappears. If it explains why your code fails after you have attempted the problem, it can support learning. If it writes the entire assessed program before you understand the logic, the productive struggle has been removed.

The same applies to reading. Asking AI to help identify concepts or questions to watch for can prepare you to read more purposefully. Asking it to summarise every assigned paper so you never engage with the original argument removes the close reading, evaluation and source judgement that academic work requires.

4. Could You Explain and Defend the Result Without AI?

Intellectual ownership is more demanding than changing a few words in AI-generated text. You own the work when you understand the reasoning, know where the evidence came from, can explain the choices and can defend the result when questioned.

Close the AI window and test yourself. Why is this argument persuasive? Why did you choose this method? What evidence supports the claim? What assumptions are being made? What would change your conclusion?

If you cannot answer those questions, the AI has probably done more than assist. This is also where verification matters. Language models can produce convincing falsehoods, invented references and distorted summaries, so any factual, numerical or cited claim that matters to your work needs independent checking. Vappingo’s guide to AI hallucinations in academic writing explains why fluent output and reliable evidence are not the same thing.

5. What Is the Smallest Useful Role AI Can Play?

The choice is rarely limited to “use AI” or “do everything alone”. The better question is how little delegation will give you the help you need while leaving the important thinking with you.

For many student tasks, a smaller role produces better learning than asking for the finished answer. AI can act as a tutor that explains one step and checks your understanding, a questioner that probes your argument, a critic that identifies weaknesses in a draft, a practice partner that generates new problems, or an organiser that helps you sort material you already understand.

It can also support research without becoming the source. For example, you can ask for alternative search terms, names of concepts to investigate or questions to pursue, then move into library databases, original papers and authoritative sources yourself.

This approach also gives you a cleaner way to use AI when policy permits assistance but expects you to remain the author. UCL’s Category 2 guidance, for example, describes GenAI as an assistive tool that should support the student’s work rather than complete the assessment on the student’s behalf.

Turn the Five Answers Into One of Four Decisions

After the five-question check, you should be able to place the task into one of four broad decisions.

The categories are not universal university policy. They are a decision aid for thinking through the role AI should play after you have checked the actual rules that apply to your work.

When AI Is Allowed but Still a Poor Choice

Permission answers one question: May I use AI here? It does not answer whether AI is a good learning decision.

A language student can be allowed to use AI for practice and still weaken their progress by asking it to rewrite every sentence. A programming student can be allowed to use a coding assistant and still become dependent on generated solutions they cannot reproduce. A student preparing for an exam can be allowed to use AI freely during revision and still learn very little if every difficult question is answered before they attempt it.

Usefulness depends on the goal. If the task is repetitive administration around the learning, automation can be sensible. If the task is the learning, handing it away can leave a gap that only becomes visible later, often in an exam, viva, seminar or more advanced module.

The goal is therefore not maximum AI use. It is deliberate AI use that leaves you more capable at the end of the session.

Worked Example: Should AI Help With a Critical Essay?

Imagine an assessment asks you to write a 2,000-word critical essay evaluating whether social media has improved political participation. The guidance permits AI for brainstorming and language support but states that the argument and substantive content must be the student’s own.

You are tempted to prompt:

Question 1: Is AI allowed here?

Yes, but only in a limited assistive role. The assessment instructions rule out delegating the substantive argument.

Question 2: What are you supposed to demonstrate?

You need to build and defend a critical position using evidence. Selecting the thesis, deciding how the argument develops and weighing conflicting evidence are central to the task.

Question 3: What thinking would you lose?

If AI creates the thesis and structure before you have analysed the problem, you lose the work of deciding what matters, identifying tensions in the evidence and choosing a defensible position.

Question 4: Could you defend the result?

At this stage, no. You have not yet established why the proposed argument is stronger than the alternatives or checked the suggested sources against the original research.

Question 5: What is the smallest useful AI role?

Develop your own initial thesis and supporting points first. Then use AI as a critic rather than an author:

You can also use the tool to suggest search vocabulary for concepts that emerged from your own analysis. Any source that enters the essay should then be found, read and verified in the original academic literature rather than cited from an AI response.

How the Decision Changes Across Common Student Tasks

The five questions are more useful than a fixed list because the same AI action can be appropriate in one context and inappropriate in another.

None of these examples overrides the rules of a particular assessment. Their purpose is to show how delegation changes the learning consequences of the same tool.

Frequently Asked Questions About Using AI for University Work

The Rule to Remember

AI fluency starts before you type a prompt. The first skill is deciding whether the tool belongs in the task at all and, if it does, which part of the work should remain yours.

Check the rules. Identify what the task is meant to teach or assess. Protect the thinking you need to practise. Make sure you can explain and defend the result. Then give AI the smallest role that genuinely helps.

Once you have made that decision, continue with the full Vappingo Student AI Loop: DECIDE, EXPLAIN, TEST, VERIFY, USE.