Learning with AI
For the first time in history, everyone with a phone has a private tutor — one that never sighs, never judges, never runs out of patience, and charges nothing to start. Used well, this may be AI's single most life-changing ability. Here's how to use it well.
The infinitely patient tutor
Think about what makes human tutoring so effective — and so expensive. A tutor explains things at your level, notices when you're lost, re-explains a different way, and lets you ask the "stupid" questions you'd never ask in a full classroom. AI does all of this, on demand, at 2 a.m. if that's when you study.
The core moves:
- Set the level. "Explain inflation like I'm 12." "Explain it assuming I know basic accounting." "Now the university-level version." Same topic, three depths — you choose the rung of the ladder.
- Re-ask without embarrassment. This is the quiet superpower. Nobody asks a teacher to explain the same thing a fourth time. With AI you can say "I still don't get it — try again, slower, different angle" as many times as you need. No sighing. No judgement. For shy learners this changes everything.
- Analogies on demand. "Explain how vaccines work using a cricket analogy." Weird requests like this genuinely work, and a good analogy sticks for years.
- Worked examples. "Show me a percentage problem solved step by step, explaining the why of each step — then give me a similar one to try myself."
- Practice questions with feedback. Answer, get corrected, understand the mistake, try again. That loop — not reading — is where learning actually happens.
- Summarise, then quiz. Paste (or photograph) a chapter: "Summarise the key points, then quiz me on them one question at a time." Instant revision session.
- Learning plans. "Build me a 30-day plan to learn spreadsheet basics, 20 minutes a day, with a small practice task each day." Structure is half the battle, and AI is excellent at structure.
A textbook gives everyone the same page 47. A tutor watches your face and adjusts. The mistake most people make with AI is using it like a textbook — one question, one answer, close the tab. The value is in the back-and-forth: "slower", "why?", "give me another example", "okay, now test me". Ten exchanges beat one perfect question.
The tutor loop
Effective sessions follow a rhythm that real tutors use instinctively. You can ask the AI to run it deliberately:
You don't have to manage this loop by hand. Say it once at the start — that's what the first prompt card below does — and the AI will run the whole rhythm for you.
Three power techniques
These three moves separate casual users from people who genuinely learn faster with AI:
- Teach-back. "I'll explain this topic to you in my own words. Point out everything I got wrong or fuzzy." Explaining something is the fastest way to discover you don't understand it — and AI is a tireless audience. Teachers call this the protégé effect; you can just call it cheap and effective.
- Socratic mode. "Don't give me answers. Ask me guiding questions and let me work it out myself. Only reveal the answer if I'm stuck twice." This turns AI from an answer machine into an actual teacher. Especially good for maths and logic, where being handed the answer teaches you nothing.
- Explain this photo. Snap the confusing textbook page, the diagram, the handwritten formula on the whiteboard — upload it and ask. "What is this diagram showing? Walk me through it arrow by arrow." The camera turns any printed page into a conversation.
Act as my personal tutor for this topic: [topic]. My current level: [beginner / know a little / studied it years ago]. Rules for this session: 1. Explain one small piece at a time, with an everyday analogy. 2. After each piece, ask me one question to check I understood. Wait for my answer. 3. If I'm wrong, don't just correct me — show me where my thinking slipped. 4. Every 3rd question, make it slightly harder. 5. Keep answers short. This is a conversation, not a lecture.
Language practice that talks back
Language learning may be the best fit of all, because what learners lack most is a patient conversation partner. AI will happily be one — in text, or out loud in voice mode while you walk (more on voice in Images, Voice & Video).
The trick is to ask for gentle correction, not a lecture after every sentence:
Let's have a conversation in English so I can practise. I'm an intermediate learner; my first language is Bangla. Rules: 1. Talk with me about everyday topics — ask me questions too. 2. When I make a mistake, first reply naturally to what I said, THEN show the correction in one short line: (Better: "..."). 3. Use vocabulary slightly above my level and explain any hard word in brackets. 4. At the end, when I say "report", list my 5 most frequent mistakes with examples. Start by asking about my day.
Swap the languages and the same prompt teaches a foreigner Bangla, or you Arabic, Japanese, German — any language the model knows, which is most of them.
Four learners, four sessions
To show the range, here's the same tool in four very different hands:
| Learner | What a session looks like |
|---|---|
| A student stuck on maths | Photographs the algebra problem: "Don't solve it — ask me questions until I can solve it myself." Then: "Give me 3 similar problems and check my answers." |
| An adult improving their English | Ten minutes of voice-mode conversation on the commute, with the correction rules above. "Report" at the end shows the week's recurring errors. |
| A professional learning Excel | "30-day plan, 20 min/day, for spreadsheet basics — with one tiny practice task daily using my own shop's sales numbers." Stuck on day 9? Paste the error, get unstuck. |
| A grandparent learning the smartphone | "Explain very slowly, one step at a time, how to make a video call on WhatsApp. Big steps, no jargon. Wait for me to say 'done' after each step." |
The honest limits
AI is a wonderful tutor and a risky authority. Keep three cautions taped to the fridge:
- Verify facts for serious study. For exams, professional certifications, medicine, law — treat AI explanations as a brilliant study partner's notes, not the syllabus. It can mangle dates, misstate formulas, and invent citations and links that look real but aren't (the full story is in Hallucinations). Cross-check anything you'll be graded or judged on against the official source.
- It's a tutor, not an accredited source. "AI said so" convinces no examiner, no client, no doctor. Use it to understand the material faster; cite the textbook.
- For children: supervise, but don't miss the opportunity. A patient tutor for every child is an enormous, genuine win — homework help, endless "why?" answering, reading practice. But chatbots aren't designed as children's products: sit nearby, use family accounts where offered, and teach kids early that AI can be wrong. The same double-checking habit you're learning is the one they need.
There's a trap here: AI can produce your essay, your assignment, your answers — and quietly rob you of the learning. The test is simple: after the session, can you do it without the AI? If yes, it tutored you. If no, it replaced you — and exams, interviews, and real life have a way of finding out. Socratic mode and teach-back exist precisely to keep you on the right side of that line.
Key takeaways
- AI is an infinitely patient tutor: set your level, re-ask endlessly, demand analogies, worked examples, quizzes, and 30-day plans.
- Run the tutor loop — explain → check → practice → feedback → harder. Ten exchanges beat one perfect question.
- Power moves: teach-back ("correct my explanation"), Socratic mode ("no answers, only questions"), and photographing whatever confuses you.
- Language practice with gentle-correction rules turns dead time into speaking time.
- Verify anything that matters against official sources — it's a tutor, not an accredited authority. For kids: supervise, and teach the double-check habit.
- The goal is that you can do it without the AI afterwards. If not, it wasn't tutoring.