Learn 10 research-backed methods and prompts to use AI for deeper learning, critical thinking and skill development.

The thinking sandwich method: Always start by thinking independently first, use AI to critique your work, then selectively integrate feedback..
10 practical techniques with ready-to-use prompts: Including the Feynman Technique for mastering concepts, Socratic questioning for deeper understanding, metacognitive tools for self-regulated learning, and deliberate practice frameworks for skill development.
Studies show mindless AI use hinders critical thinking and memory retention, while strategic use enhances learning. This guide provides the specific prompts and methods to use AI as a thinking partner, not a replacement.

Practical methods and prompts for learning

Research studies shows that how we use AI chatbots matters more than whether we use them. Studies from institutions including the Wharton School and MIT have found that mindless engagement with chatbots can hinder critical thinking and memory retention. However, strategic use can enhance learning, deepen understanding, and accelerate skill development.

This guide presents research-backed methods and ready-to-use prompts for transforming AI chatbots from answer machines into genuine thinking partners. The core principle throughout:

Always do your own thinking first, then use AI to challenge, refine and expand your understanding.

Essential principles

Always start without AI. Struggle first to build genuine understanding.
Remain sceptical. Verify factual claims; AI can hallucinate confidently.
Request questions, not answers. The best prompts ask AI to probe your thinking, not replace it.
Reflect regularly. Ask yourself if AI use is helping or hindering your long-term goals.

Part 1: The thinking sandwich method

The most important rule for learning with AI is to structure your work like a sandwich, where your thinking forms both the foundation and the crown of the process.

The 3 layers

Bottom layer (Your raw material)
Start by producing something on your own. Draft an argument, outline a presentation, or articulate your understanding of a concept. This struggle builds neural pathways and genuine comprehension.
Middle layer (AI support)
Ask the AI chatbot to critique your work, identify blind spots, challenge assumptions, or suggest different angles. Reflect carefully on this feedback.
Top layer (Integration)
Selectively incorporate feedback into your original work. Decide for yourself what strengthens your thinking and discard what doesn’t. The final product remains authentically yours.

Sample prompt for middle layer:

I’ve drafted my initial thoughts on [topic]. Please critique my reasoning by: (1) identifying logical gaps or weak points in my argument, (2) pointing out assumptions I may not have examined, (3) suggesting perspectives I haven’t considered, and (4) noting where my evidence is thin. Don’t rewrite anything. Just provide critical feedback I can use to strengthen my own thinking.

Part 2: Two learning modes

Directive prompts give explicit instructions (e.g., “Write a poem about X in the style of Y”), while non-directive prompts offer open-ended guidance, encouraging exploration and self-discovery (e.g., “Tell me about your feelings on X”). Directive prompts are specific and task-oriented, good for clear outcomes, while non-directive prompts foster deeper, self-directed growth and creativity

Different learning situations call for different types of AI interaction. Understanding when to use directive versus non-directive prompts dramatically improves your learning outcomes.

Directive mode: The critical reviewer

When to use: You have a tangible product, a finished draft, developed idea, or completed analysis, and want expert-level feedback.

Directive mode prompt:

Act as a critical reviewer. Evaluate the clarity of my argument, the logic of my structure, and the persuasiveness of my evidence. Point out weaknesses or gaps, and suggest ways to improve the flow and coherence. Do not rewrite the text yourself. Focus only on critical feedback.

Non-directive mode: The Socratic tutor

When to use: Your ideas are half-baked, you’re exploring new territory, or you want to develop your own thinking without being steered toward a particular conclusion.

Non-directive mode prompt:

Never tell me directly what to fix and avoid strongly imposing your own vision. Instead, point neutrally and vaguely to areas that might need more consideration. For example, raise potential ambiguities, confusing phrasing, or ideas that could be strengthened. If you see a concrete issue, don’t write ‘Argument X is weak, you need to add Y.’ Instead, write something like ‘Are there any weak spots in argument X? What points could be criticised by a sceptic?’

Socratic sparring partner prompt:

I have an idea about [topic], but I haven’t fully developed it yet. Act as a Socratic sparring partner: ask me questions that challenge my assumptions, clarify my goals, and make me consider aspects I haven’t articulated yet. Do not provide any answers, opinions or suggestions. Stay as neutral as possible. Your role is only to provoke my thinking and help me understand my own ideas better.

Part 3: The Feynman technique with AI

Richard Feynman believed you only truly understand something if you can explain it simply. This technique, supercharged with AI feedback, is one of the most powerful learning methods available.

The 4-step process

Learn the basics: Study the concept from primary sources first.
Explain it simply: Write or speak your explanation as if teaching a beginner.
Identify gaps: Use AI to analyse your explanation and reveal what you missed.
Refine and simplify: Return to source material, fill gaps, and try again.

Feynman technique prompt:

I’m going to explain [concept] as if teaching someone with no background in this area. After I finish, please: (1) identify any parts of my explanation that are unclear, overly complex, or use jargon without defining it, (2) point out any factual errors or misconceptions, (3) highlight gaps—important aspects of the concept I didn’t cover, and (4) suggest where I could use better analogies or examples. Don’t fix the explanation yourself—help me see where I need to deepen my own understanding.

Reverse Feynman prompt:

Ask me to explain [concept] as if I’m teaching it to a 12-year-old. After each of my responses, evaluate whether my explanation was clear, accurate, and simple enough. Ask follow-up questions to probe areas where I might be glossing over complexity or hiding behind jargon. Continue until you’re satisfied I truly understand the concept at a fundamental level.

Part 4: Metacognitive learning prompts

Metacognition, thinking about your own thinking, is the foundation of effective learning. Research shows that AI chatbots with metacognitive prompting significantly enhance self-regulated learning abilities.

Planning phase:

I want to learn [topic/skill]. Help me create a learning plan by asking me questions about: what I already know about this topic, what specific outcomes I want to achieve, how much time I can dedicate, and what learning resources I have available. Based on my answers, help me structure a realistic study plan—but make sure I do the planning work, not you.

Monitoring phase:

I’ve been studying [topic] for [time period]. Ask me questions to help me assess my progress: What concepts do I feel confident explaining? Where am I still confused? What strategies have worked well? What hasn’t been effective? Help me identify whether I’m on track with my learning goals without giving me the answers directly.

Reflection phase:

I’ve just completed learning about [topic]. Help me reflect deeply by asking me: What was the most challenging part and why? What would I do differently if starting over? How does this connect to other things I know? What questions do I still have? What’s the most important insight I gained?

Self-regulated learning cycle:

Act as my learning coach. After each study session I report to you, ask me three types of questions: (1) Planning questions before I start: what’s my goal, what strategy will I use? (2) Monitoring questions during: how confident am I, what’s working? (3) Reflection questions after: what did I learn, what will I change? Track my responses over time and help me notice patterns in my learning.

Part 5: Critical thinking development

Critical thinking requires practice examining arguments, challenging assumptions, and considering multiple perspectives. These prompts turn AI chatbots into rigorous thinking partners.

Assumption destroyer:

I believe [statement or position]. Help me examine this belief by: (1) identifying the key assumptions underlying my position, (2) for each assumption, asking ‘what if this weren’t true?’, (3) presenting the strongest possible counterargument, and (4) identifying what evidence would change my mind. Don’t try to change my view—help me stress-test it rigorously.

Devil’s advocate:

I’m developing an argument that [your position]. Play devil’s advocate: argue against my position as persuasively as possible. Then, ask me to defend my position against your best counterarguments. Continue this back-and-forth debate until my argument has been thoroughly stress-tested. At the end, identify which of my rebuttals were strong and which need more work.

Cognitive bias scanner:

I’ve made a decision about [situation]. Analyse my reasoning for potential cognitive biases: confirmation bias, anchoring, availability heuristic, sunk cost fallacy, etc. For each bias you identify, explain how it might be affecting my thinking and suggest specific questions I should ask myself to counteract it. Be specific about where in my reasoning each bias appears.

Multi-perspective analysis:

I’m thinking about [issue or decision]. Help me see this from multiple perspectives by asking me to consider: (1) how would someone who disagrees with me view this? (2) how would this look in 10 years? (3) what would an expert in [relevant field] notice that I’m missing? (4) what are the second and third-order consequences I haven’t considered? Ask me these one at a time and probe my answers.

Historical perspective challenge:

Critique my idea about [topic] from two contrasting perspectives: first, as an ancient philosopher skeptical of modern assumptions, then as a futurist considering where current trends lead. Ask me questions that each perspective would raise, and help me see blind spots that my contemporary viewpoint might create.

Part 6: Creative thinking enhancement

Creativity flourishes when we make unexpected connections and escape conventional thinking patterns. These prompts push you beyond predictable responses.

Cross-domain insight generator:

I’m working on a problem in [your field]: [describe problem]. Ask me to consider: what parallel problems exist in [unrelated field]? How have those fields solved similar challenges? What principles or metaphors from [unrelated field] might apply here? Push me to make at least three unexpected connections between domains.

Metaphor finder:

I’m trying to explain [complex concept] to [target audience]. Help me find powerful metaphors by asking me about the core essence of this concept. What is it really about at the deepest level? Then challenge me to generate metaphors from everyday life, nature, sports, cooking, or other familiar domains. For each metaphor I suggest, probe where it works and where it breaks down.

Constraint creativity:

I need creative solutions for [challenge]. Apply creative constraints to push my thinking: what if I had only 10% of the budget? What if I had to solve this in one day instead of one month? What if the obvious solution were impossible? What if I had to make this work for a completely different audience? Ask me to generate solutions under each constraint before removing them.

Idea synthesis:

I have several incomplete ideas about [topic]: [list your ideas]. Help me combine them into something more powerful by asking: what do these ideas have in common at a deeper level? Where do they contradict each other productively? What new possibility emerges if we combine the best elements of each? Push me to synthesise rather than simply choose between them.

Part 7: Skill development prompts

Deliberate practice, focused effort on specific weaknesses with immediate feedback, accelerates skill development. These prompts structure deliberate practice conversations.

Gap analysis:

I want to become proficient at [skill]. First, ask me to describe what an expert in this skill can do. What separates masters from beginners? Then ask me to honestly assess where I am on that spectrum and why. Help me identify the 2-3 specific sub-skills that would create the biggest improvement if I focused on them. Don’t tell me what to work on. Help me figure it out through questioning.

Deliberate practice designer:

I want to improve my [specific skill or sub-skill]. Help me design a deliberate practice session by asking: what specific aspect should I focus on? What would success look like? How will I get feedback? What makes this challenging enough to create growth but not so hard I can’t improve? How long should each practice session be? Guide me to create a concrete practice plan.

Skill transfer:

I’ve developed skill at [existing skill] and want to apply it to learning [new skill]. Ask me questions to help identify: what principles transfer directly? What habits or mental models from [existing skill] might actually interfere with learning [new skill]? What do experts in [new skill] do differently from experts in [existing skill]? Help me leverage my existing competence without being limited by it.

Part 8: Knowledge verification and testing

Testing yourself is one of the most effective learning strategies. These prompts help you assess your knowledge accurately.

Socratic examination:

Test my understanding of [topic] using Socratic questioning. Start with foundational questions and progressively increase difficulty based on my answers. If I answer correctly, probe deeper. If I struggle, ask clarifying questions rather than giving answers. After 10-15 questions, summarise what areas I understand well and where I have gaps. Be rigorous. Don’t accept vague answers.

Application challenge:

I’ve been learning about [concept/theory]. Test my understanding by presenting me with novel scenarios where I need to apply this knowledge. Don’t give me textbook problems. Create realistic, messy situations where the correct approach isn’t obvious. After each scenario, ask me to explain my reasoning before evaluating whether I’ve applied the concepts correctly.

Prediction exercise:

I’m studying [topic]. Before I research further, ask me to make specific predictions about [aspect of topic]. Record my predictions. Then, as we discuss the actual answers, help me analyse: where was my intuition correct? Where was I wrong, and why? What does the pattern of my errors reveal about misconceptions I should correct?

Quick reference decision tree

Use this decision framework before engaging with AI on any learning task:

Is this a skill I want to develop long-term?
If yes, always start without AI and use the sandwich method.
Do I have a formed product (draft, outline, idea)?
If yes, use directive mode for critique. If no, use non-directive mode.
Am I trying to understand something deeply?
Use the Feynman technique. Explain first, then get AI feedback.
Am I preparing for a test or assessment?
Use Socratic examination and Application challenge prompts.
Am I working on creative problems?
Use Creative thinking prompts with cross-domain connections.
Is this routine work that won’t impact my development?
AI assistance is appropriate, but verify important claims.

The goal is not to resist AI, but to use it wisely as a tool that amplifies your thinking rather than replacing it.

Recursos

Your brain on ChatGPT (Artículo)
Without Guardrails, Generative AI Can Harm Education (Artículo)

How to Use AI to Get Smarter (and not Dumber)