Use AI without handing over your thinking

Before: write your own starting point

Before opening an assistant, spend a few minutes writing what you already think. For a study question, note your tentative answer and where you are uncertain. For a work decision, list the options and the information you still need.

This gives you a reference point. Without one, a fluent response can become the default account simply because it arrived first. Your initial view need not be correct; the purpose is to make your own assumptions available for comparison.

Check the rules for the task

For assessed work, read the assignment’s AI instructions and your institution’s current policy. Permission to brainstorm does not necessarily include permission to generate submitted text. If the rules are unclear, ask the course team before using generated material in an assessment.

At work, check approved tools and information-handling rules. Do not assume that removing a name makes a document safe to share; a combination of details can still identify someone. UNESCO’s guidance on AI in education and research provides broader discussion of human agency and privacy, but your actual task rules must come from the responsible institution.

During: request criticism or practice

Ask the assistant to examine a defined piece of thinking rather than produce a final verdict. It can suggest a counterargument, identify a possible ambiguity, or offer a practice question. Treat those suggestions as candidates to investigate.

For example, suppose your position is that recorded lectures always improve learning. Ask what the word “always” requires you to establish, what other explanations might exist, and what evidence could challenge the claim. A suggested objection is useful only if you can understand and assess it.

Here is my own argument. Identify two assumptions and one plausible counterargument. Do not rewrite it or invent citations. Explain what evidence would help me decide whether each concern is relevant.

Keep evidence separate from generated explanation

A generated citation can look convincing without being real or relevant. Open a source, confirm its authorship and publication details, and read the passage supporting the claim. A title alone is not evidence that the study found what the assistant says.

Keep a small record of the claim, the source and your interpretation. Distinguish findings from your inferences. If you cannot obtain the source, do not present the assistant’s description as though you personally verified it. A second generated answer is not a substitute for the original evidence.

After: explain it without the chat

Close the answer and explain the central idea in your own words. Can you give a new example? Can you say why an alternative is weaker? If you can only repeat the wording, return to the underlying material.

This is a practical self-check, not a formal test of learning. It helps identify where more reading or practice would be useful. For a workplace decision, the equivalent is being able to explain your reasons and uncertainties to someone affected by it.

Keep a record of the help you used

Where disclosure is required, record the tool, date, purpose and how you used its output. Keep drafts or notes when your institution or workplace expects them. Do not claim that work was entirely unaided if that would be misleading.

You remain responsible for checking the final result and deciding whether it belongs in your work. AI assistance can provide a useful challenge or starting point. The aim is to finish with a clearer understanding that you can explain, question and defend yourself.

Sources and further reading

The exercises and examples are Virel AI’s practical suggestions. These resources provide supporting guidance and broader context.

• UNESCO: Guidance for generative AI in education and research