Use AI to support practice, then check what you can still explain and do independently.
The most useful learning question is not simply which task a chatbot can complete. It is what you need to understand to set the task, judge the result, and act on it responsibly. A generated answer may help you finish an assignment while leaving you unable to solve a similar problem tomorrow. Equally, a well-designed exercise can use AI to expose gaps in your reasoning and give you more opportunities to practice. Building skills in this environment means deciding when you want a finished output and when the struggle of producing it is part of what you need to learn.
A better answer is not always better learning
A 2025 PNAS field experiment involving nearly 1,000 high-school mathematics students in Turkey compared two GPT-4-based tutors with a control group. The basic assistant improved performance during assisted practice, but its group performed worse when the tool was removed. A tutor designed with teacher-provided safeguards largely mitigated that negative effect. This was one educational setting and model configuration, not evidence that every AI tutor harms learning or that safeguards always succeed.
The practical distinction is between success with a tool and the knowledge you retain without it. A sensible learning plan checks both.
Four skills to practice together
UNESCO’s 2024 student competency framework, whose web summary was updated in January 2026, covers human agency, ethics, AI techniques and applications, and system design. It treats students as people who evaluate and shape technology, rather than merely operate interfaces. Building on that perspective, AI Affect suggests four connected habits.
- Subject knowledge: Learn enough about your field to recognize missing steps, wrong assumptions, and misleading terminology. Choose a real problem and explain the relevant concepts in your own words.
- Problem framing: State the objective, audience, constraints, and what would count as success before asking for an answer. Practice distinguishing the decision you need to make from the information you are missing.
- Evidence checking: Trace an important claim to its supporting material. Check that a linked source exists, says what is claimed, and applies to your situation. Separate a source’s findings from your interpretation.
- Communication and judgment: Explain a recommendation and its uncertainty to another person. Identify who could be affected and who must approve or take responsibility for the outcome.
None of these is a promise of permanent protection from labor-market change. Together, they give you a stronger basis for using unfamiliar tools and evaluating your own progress.
Use a practice loop you can inspect
Start with a small problem and attempt it without assistance. Write down where you become stuck. Then use AI to request a hint, a counterexample, or a critique of a particular step. Avoid requesting the final solution immediately when learning the method is your goal.
After checking the explanation against reliable course material, close the tool and solve a related problem. Explain the reasoning aloud or write a short account from memory. If you still cannot do it, record the gap and return to the underlying concept. This is a suggested study routine, not a tested guarantee of improved grades.
Make progress visible
Keep a small portfolio with the original task, your first attempt, the help you used, and the final result you can explain independently. Include an error you identified and how you checked it. For an essay, that might be a source that did not support a claim. For a spreadsheet, it might be an incorrect formula caught by reconciliation.
Follow your school’s or employer’s rules about permitted assistance and disclosure. Choose tools appropriate to the information you are allowed to share. The aim is to finish a learning session with more usable knowledge and clearer judgment, alongside whatever document or answer you produced.