How to use AI to study without stopping yourself from thinking: five checks to carry out before accepting an answer

The quickest answer lets you skip the part where you learn

You need to understand a concept, solve an exercise, prepare a presentation or write a text. You open an AI assistant, make a request and, within seconds, receive a well-structured, fluent response that seems better than anything you could have produced yourself.

But a flawless answer can also mask the fact that you haven’t learnt anything.

Before accepting what appears on screen, there are questions you should ask yourself first: can you use that tool for this task? Do you know where the information comes from? Do you understand it well enough to defend it without reopening the app?

Knowing how to use AI for studying isn’t just about pressing a button. It’s about knowing which decisions should remain yours. The five checks below will help you distinguish when the tool is deepening your understanding and when it’s simply replacing the work you needed to do yourself.

Check 1 — The activity’s rules come before the prompt

The first step isn’t to choose a tool or write a prompt. It’s to check whether you’re allowed to use artificial intelligence for that specific activity.

Each school, subject, teacher and type of activity may set different conditions. A use that is permitted for revising a topic on your own may not be allowed in an exam, an assignment or a project that is subject to assessment. The absence of an explicit ban does not equate to authorisation.

Before opening any assistant, answer these three questions:

1. Is the activity for personal study or is it assessed?
2. Do the instructions permit, restrict or prohibit the use of AI?
3. What specific uses has the teacher authorised?

Institutional regulations take precedence

At UDIT, the use of artificial intelligence in an academic activity must comply with current regulations and the instructions given by teaching staff. Before using it, check the course instructions or consult your lecturer.

This distinction is fundamental to any discussion of academic integrity and artificial intelligence: citing a tool may highlight its involvement, but it does not grant permission to use it. And if you’re wondering how to cite ChatGPT or any AI assistant in an assignment, the answer begins before the citation: first, confirm that you’re allowed to use it.

Check 2 — Take an initial step that the AI cannot do for you

Before consulting a tool, spend a few minutes producing something of your own: a hypothesis, an outline, a question, a sketch, a calculation, a list of uncertainties or a provisional explanation. It doesn’t have to be perfect. It just needs to help you identify where you’re starting from.

The ‘zero draft’ technique

Write down your thoughts before opening any assistant:

• What do I already understand about this topic?
• Where do I have doubts?
• What solution would I try?
• What information do I need to find?

Using that draft as a reference, you can use AI to spot gaps in your reasoning, ask for counterexamples, check your interpretation against the facts, or request feedback on your outline.

Without a starting point of your own, it’s difficult to tell the difference between having developed an idea yourself and having been given a ready-made one. When someone is looking for ways to learn with artificial intelligence, the most useful answer isn’t what to ask the assistant, but what to ask yourself before consulting it.

Checkpoint 3 — Ask for help to get past the difficulty, not to make it disappear

The educational value of artificial intelligence for students depends on the role you assign to it. Asking for a supplementary explanation is not the same as asking it to write the assignment for you.

Purpose test

Before continuing, complete this sentence:

‘I am using AI to ________, but the decision or task that remains mine is ________.’

If the second part is left blank, the use is probably taking on too much of a substitute role. The responsible use of AI in education is not measured by the tool you choose, but by the role you assign to it.

Check 4 — A convincing answer still requires evidence

A well-written piece may contain fabricated data, non-existent references, made-up quotes, incorrect dates, out-of-date legislation, vulnerable code or conclusions lacking context. The fluency of the text does not guarantee the accuracy of its content.

If you’re looking for ways to verify AI-generated information, follow this five-step verification process:

Five-step verification protocol

1. Extract the claims. Isolate specific data, dates, names, figures, laws, concepts and references.
2. Find the original source. Prioritise official documents, scientific articles, regulations, datasets or technical documentation.
3. Check the context. Review the date, authorship, geographical scope, version and any updates.
4. Cross-check where necessary. Look for a second independent source if the claim is current, controversial, technical or legal.
5. Test what can be tested. Run the code, repeat the calculation, open the link, validate the model or consult the full regulation.

Hierarchy of evidence

A generated response can be a starting point. It should not automatically become the evidence underpinning your work. When someone wonders whether using ChatGPT or any other AI for studying is a good idea, the answer depends on what you do next with the output you receive.

Check 5 — Make it clear what the tool did and what you did

Where its use is expressly permitted, you should be able to explain: which tool you used, what you used it for, what information you entered, what you received, which parts you kept, which you discarded, which sources you verified, and what decisions you made.

The exact way of citing or acknowledging the AI depends on current regulations, the teacher’s instructions and the required citation system. This is important: the issue of AI and academic plagiarism cannot be resolved by a universal citation format. It is resolved through authorisation, transparency and the ability to defend your work.

Minimum record of the process

Keep the following for the duration of the activity:

• Your own initial draft
• Tool and date of use
• Relevant prompts
• Responses used
• Sources consulted for verification
• Decisions and changes made
• Final version

Defence test

Before handing in any work, answer honestly:

• Can I explain the content in my own words?
• Do I know where each statement comes from?
• Can I reproduce or explain the steps?
• Do I still have a version prior to the AI’s intervention?
• Do I take responsibility for the result?

If you cannot explain, verify and defend the result, you should not yet present it as your own learning.

The same practice can take on different meanings depending on the context

Not all contexts work in the same way. What may be useful for personal study could constitute cheating in an assessed assignment.

Five responsibilities that must remain recognisably yours

Regardless of the tool you use, there are five elements of learning that should not be lost sight of:

1. The question: what do you want to understand or resolve?
2. Selection: what information, references or alternatives you consider valid.
3. The criteria: why you choose one option and rule out others.
4. Verification: how do you know that the result works or is true?
5. Justification: how you explain the process and take responsibility for the result.

These five responsibilities apply to any discipline. In graphic design, the concept and visual criteria. In programming, the architecture and understanding of the code. In data science: the selection of metrics, biases and reproducibility. In fashion: cultural research and materiality. In product and interior design: measurements, regulations and feasibility.

AI can provide support in each of these areas. What it cannot do is turn an automated response into your own judgement.

Before uploading a file, check that you have the right to share it

This point is often overlooked: privacy, confidentiality and intellectual property are also part of the academic use of AI.

Do not enter personal data, records, private emails, other people’s work, confidential briefs, proprietary code, restricted datasets, exams, unpublished research or protected materials without authorisation.

Quick checklist

• [ ] Is this content mine?
• [ ] Do I have permission to share it?
• [ ] Does it contain personal information or information about third parties?
• [ ] Does the platform store the data or can it use it for training purposes?
• [ ] Does the activity allow the use of external services?

The convenience of pasting an entire document does not remove obligations regarding privacy, confidentiality or intellectual property.

AI cuts across different disciplines, but does not remove the responsibility of each one

The criteria for using artificial intelligence are not limited to technology-related degree programmes. In Design, authorship and the ability to justify visual decisions are important. In Animation and Digital Art, artistic direction and rights over materials. In Video Games, the integration and testing of each output. In Full-Stack Development, the understanding and auditing of code. In Data Science, biases, traceability and reproducibility.

UDIT, as a university specialising in Design, Innovation and Technology, approaches AI from the perspectives of applied use, critical understanding, technological creation, ethics and the professional context of each field. The University’ s own degree in Applied Artificial Intelligence complements this training in a cross-disciplinary manner for students on any degree programme.

When comparing universities, ask what ‘using AI’ means there

If you’re considering where to study, a university’s artificial intelligence policy also says a great deal about its academic approach. These questions may help you during an open day or an admissions interview:

1. Are there any public regulations on artificial intelligence?
2. Is the general rule one of permission, prohibition or left to the discretion of individual lecturers?
3. How is it communicated which uses are authorised for each activity?
4. Are students taught how to verify AI results?
5. Is there a requirement to declare the tools and prompts used?
6. How is authorship protected in creative projects?
7. Is only the outcome assessed, or is the process assessed as well?
8. What supplementary training in AI does the university offer?

Don’t just ask which tools you’ll be able to use. Ask what criteria you’ll learn to apply when a tool produces something for you.

Frequently asked questions

Can I use ChatGPT or another AI tool to study?

For personal study, you can use it to ask for explanations, practise or identify areas of uncertainty, provided you do not share confidential information and you verify its answers. For assessed work, check first whether your lecturer and the regulations permit that specific use.

Is using artificial intelligence in an assignment considered plagiarism?

It depends on the rules of the assignment and how the tool is used. Presenting generated content as your own, concealing its use, or using it when it is prohibited may constitute academic misconduct. Simply acknowledging its use is not sufficient if the assignment did not authorise its use.

Is it enough simply to cite ChatGPT?

No. Citing the source provides transparency, but it does not replace authorisation. Furthermore, claims must be supported by original and verifiable sources. The exact way to acknowledge or cite an AI depends on institutional regulations and the lecturer’s instructions.

How can I check whether an AI response is correct?

Identify specific claims, look for the original sources, check the date and context, cross-check relevant points and verify whatever can be verified. Do not use bibliographic references without checking them: generative systems can invent authors, titles or quotations.

Should I keep the prompts and responses?

Where use is authorised, keeping prompts, drafts, sources and versions helps to reconstruct the process, demonstrate the decisions made and explain the tool’s role. The specific requirements will depend on the activity and the lecturer’s instructions.

Can I submit AI-generated code?

Only when the assignment permits it and in accordance with its conditions. You must understand it, run it, test it, check its security and be able to explain every relevant decision. The fact that a piece of code works does not in itself prove that it is correct, secure or appropriate.

What information should I not enter into an AI?

Avoid personal data, confidential documents, company projects, other people’s work, exams, private code or protected materials without authorisation. Before uploading a file, check that you have the right to share it and find out how the platform will handle that information.

When comparing universities, don’t just ask whether they use artificial intelligence. Ask when they allow its use, how they teach students to verify it, which part of the process the student must carry out themselves, and how they assess whether the result demonstrates independent learning.