AI
How to study with AI without learning its mistakes
AI can summarise a lecture in seconds and write a hundred practice questions. It can also state something false with complete confidence. Here's how to get the first without the second.
What AI is genuinely good at
- Turning your material into questions, so you can test yourself.
- Explaining a concept in simpler words, or with a new example.
- Quizzing you on a topic and following up on your answers.
- Saving the hours spent formatting notes and typing cards.
Where it goes wrong
Language models produce plausible text. Usually that's correct; sometimes it isn't, and nothing in the wording tells you which. A general chatbot also answers from everything it has read, not from your lecturer's version: a slightly different definition, an outdated guideline, another country's law. In subjects marked on exact facts, those small differences cost marks. In law it can be worse: in 2023 a US court sanctioned lawyers whose brief cited cases a chatbot had invented (more in how to study law).
Four rules for studying with AI
- Give it your material, and ask it to use only that. Your slides, your book chapter, your notes.
- Ask for the source of every answer: the line and page. If it can't point to it, don't learn it.
- Use AI to make you think, not to think for you: questions, not summaries you only read.
- Check anything that surprises you against the source before it goes into your memory.
What to ask for, and what not to
| Instead of | Ask for |
|---|---|
| “Summarise this lecture.” | “Write ten questions on this lecture, using only these slides, with the slide number for each answer.” |
| “Explain heart failure.” | “Explain this paragraph from my notes in simpler words, without adding facts it doesn't contain.” |
| “What are the key cases on negligence?” | “Which cases does this reading list cite on negligence, and for what?” |
| “Answer this past exam question.” | “Mark my answer to this past question against the model answer, and tell me what I missed.” |
The pattern: keep the AI inside your material, have it produce questions and feedback rather than finished text, and ask where every answer came from.
A routine that works
After each lecture, turn it into flashcards from the slides. Review a few minutes a day with spaced repetition. When a card's answer looks off, open its source and fix or delete it. Before the exam, answer past questions closed-book and turn your mistakes into new cards. The AI saves the typing; you still do the remembering.
Built for exactly this
FlashCardz only writes what your files say, quotes the line each card came from, checks the card against it before you see it, and opens your own page in one tap. It's an AI study tool designed around the problem above.
Is it allowed?
Using AI to test yourself on your own course material is studying. Rules about AI at universities are about work you hand in; check your university's policy for that, and never submit AI-written work as your own where it's not permitted.
Questions
Can I trust AI-generated flashcards?
Only if you can check them. Use tools that work from your own material and show the source of each answer.
Which AI is best for studying?
One that works from your course material and shows its sources, rather than answering from general knowledge.
Is using AI to study cheating?
Testing yourself with AI on your own notes is studying. Submitting AI-written work may break your university's rules.