The Two Ways Students Use AI (Only One Works)
Watch two students use the same chatbot. Student A pastes the assignment, copies the answer, and moves on — fast tonight, lost in the exam hall. Student B attempts the problem first, asks AI to explain exactly where their thinking broke, gets quizzed on weak spots, and walks in able to solve it cold. Same tool, opposite outcomes.
The research on this is converging: AI as tutor accelerates learning; AI as answer-machine erodes it while feeling productive. Everything below is built for the second path — with the policy caveats to keep you out of trouble.
Workflow 1: The Explain-Quiz Loop (Any Chatbot)
The highest-value study pattern, using any capable chatbot (our comparison picks one; free tiers suffice):
- Read/watch the material first — no AI. Struggle is the point; it tells the AI what to target.
- Explain-back: 'I'll summarize this section in my own words; find the gaps and misconceptions in my summary.' This flips AI from lecturer to examiner — far more effective.
- Quiz mode: 'Ask me 5 questions on this chapter, one at a time, grade my answers, then re-ask what I missed.' Active recall with instant feedback — the study technique with the strongest evidence behind it, now free.
- Re-explain: 'Explain [concept] three ways: simply, with an analogy, and as an exam answer.' Different frames cement understanding.
Workflow 2: Notes and Revision Systems
- Lecture recordings → summaries: NotebookLM (Google) ingests your actual course PDFs and slides and answers strictly from them — grounded in your syllabus, not the internet's version. Its audio-overview feature turns notes into a podcast-style revision aid you can listen to on commutes.
- Flashcards with scheduling: generate candidate flashcards with AI, but import into Anki (free, spaced repetition) for the review schedule — the algorithm matters more than the generation.
- Past-paper drilling: paste the marking scheme and your answer; ask for the grade and the two points an examiner would have missed. Marking-scheme literacy is a real, learnable skill.
Keep your notes in a system that survives the degree — our notes comparison maps the options; Obsidian or Notion handle three years of coursework well.
Workflow 3: Writing Support (the Honest Version)
Writing assignments are where policy lines are sharpest. Safe, universally accepted uses:
- Outline critique: your thesis and structure, critiqued before you write.
- Feedback on your draft: 'mark where the argument is weak, where evidence is missing, and which paragraph a bored reader skips' — then you rewrite.
- Clarity passes on your own words: grammar and phrasing fixes are the new spellcheck — accepted at most institutions, check yours.
Never: generating submissions wholesale, 'humanizing' AI text to dodge detectors (detectors are unreliable in both directions, but procurement of fraud is fraud), or citing sources the AI gave you without reading them — fabricated citations are how students get caught even when the writing was theirs.
The Integrity Framework (Print This)
- Know your institution's policy — it varies by course, not just by university. When unclear, ask the professor; 'I want to use AI to quiz myself' is a welcome question.
- The reproduction test: could you redo this work, cold, without the AI? If yes, you learned. If no, you cheated yourself first and the institution second.
- Disclose where asked: an acknowledgments line ('used ChatGPT to quiz me on chapter 4') costs nothing and builds trust.
- Skills you cannot outsource: the exam, the interview, the job. AI-assisted learning builds those; AI-substituted work rents competence you cannot renew.
