Everyone hits a concept that won't click. Not because you're not smart enough, but because the explanation you're given assumes a path through the idea that doesn't match how you think. A different path — a different analogy, a different decomposition — is often all it takes.

AI is good at generating those alternative paths instantly. But ask it "explain X" and you'll get the most generic, textbook-average explanation it can produce. Useful as a first pass, useless when that first pass didn't work.

These 15 prompts are for when the first explanation failed. Each attacks understanding from a different angle — analogy, mechanism, contrast, or retrieval — so you can try many until one lands.

Reframe the explanation: find a path that fits you

1. The alternative analogy. Our Explain a Hard Topic prompt does analogy plus breakdown in one shot. When that analogy doesn't land, try this:

Explain [concept] with an analogy from [domain you know well — cooking, sports, music, driving]. Map each part of the concept to a part of the analogy, then tell me where the analogy breaks down and why. Keep the mapping explicit: "X corresponds to Y because..."

Where the analogy breaks is where real understanding starts. If you know what the analogy can't cover, you know what makes the concept distinct.

2. Explain at three levels. Our Explain at Three Levels prompt does this structurally — useful when you don't know which level you're stuck at:

Explain [concept] at three levels: one sentence for a 10-year-old, one paragraph for a smart peer who is new to the field, and one paragraph for an expert that includes the nuance the simpler versions omit. Then quiz me at the peer level — one question at a time.

You often don't need the expert level. But seeing what it adds tells you whether your confusion is at the core idea or the nuance.

3. From concrete to abstract.

Don't start with the definition of [concept]. Start with a specific, concrete example where it appears. Walk through that example step by step, then generalize to the definition. End with a second example that looks different on the surface but uses the same underlying principle.

Definitions-first teaching fails most learners. Concrete-first — one instance, then the pattern — matches how brains extract concepts.

4. History of the idea.

Explain [concept] by telling the story of why it was invented: what problem existed before it, what failed attempts came before it, and what insight made it work. Include the human stakes — who needed this and what couldn't they do without it.

You remember stories. "Bayes' theorem updates beliefs with evidence" is a definition. "A minister in 1763 needed to reason backwards from effects to causes and invented a way to do it" is a hook that makes the mechanism memorable.

Decompose: break the hard into small

5. First-principles decomposition.

Break [concept] into its smallest building blocks. For each block, define it in one sentence without using jargon from the other blocks. Then rebuild the concept by combining the blocks in order, stating what each addition adds. Flag which block learners usually misunderstand.

Hard concepts are bundles. If you can name the bundle's parts, you can test each part separately instead of being confused by the whole.

6. The mechanism walkthrough.

Walk me through the mechanism of [concept] as a sequence of 5 steps. For each step, tell me: what goes in, what happens, what comes out, and what would go wrong if that step failed. Use the same running example throughout.

Mechanisms beat definitions. You understand photosynthesis when you can trace the steps, not when you can recite the equation.

7. Confusion pair in parallel.

I confuse [concept A] with [concept B]. Put them in two columns: definition (one sentence), when you'd use each, and a scenario where choosing the wrong one would cause a real mistake. End with 2 quiz questions where the wrong answer is plausible.

Some concepts are hard not intrinsically but comparatively — they sit too close to a neighbor in your mind. Separating the boundary is the cure. This is the same prompt from our exam prep guide, used here for comprehension rather than point gains.

8. The abstraction ladder.

Place [concept] on a ladder: most abstract (the general principle), middle (the typical instance), most concrete (one specific case with numbers/names). Show me how to move up and down the ladder by giving me the abstract and asking me to produce the concrete, then vice versa.

Being able to move both directions — abstract to concrete and back — is the test of real understanding. One direction alone is pattern-matching.

Test yourself: retrieval reveals gaps

9. The Feynman loop. The core learning technique, structured as a prompt. Our Feynman Technique Tutor does this as a guided conversation:

I want to learn [topic] using the Feynman technique. Have me explain it in my own words, then ask one question that probes where my explanation is vague or relies on jargon. When I answer, tell me if I'm hiding behind words I don't fully understand. One question at a time until my explanation is jargon-free.

This is the single highest-value prompt on this list. If you use only one, use this. It converts passive reading into a test you'd otherwise never give yourself.

10. The Socratic probe. For when you think you understand but haven't been tested. Our Socratic Tutor prompt is designed for exactly this:

It refuses to give you answers and instead asks questions that lead you to them. Use it after you've read an explanation — it will find the place where you nod along but can't produce the next step yourself.

The manual version:

Don't explain [concept]. Instead, ask me questions about it one at a time that test whether I really understand. If my answer is vague or wrong, give me a hint that points to the gap without telling me the answer. Keep going until you've probed the three deepest layers.

11. Predict before you're told.

I'm about to learn [next part of concept]. Before you explain it, have me predict what logically should happen next given what I already know about [prior part]. Then tell me if my prediction matches reality and why or why not.

Prediction activates prior knowledge and makes the actual explanation land harder — whether you were right (confirmation) or wrong (surprise). Both encode memory better than passive reading.

12. The wrong answer autopsy.

I think [concept] means [your current understanding, possibly wrong]. Don't correct me yet — first, tell me what mistake my phrasing reveals about how I'm modelling the idea. Then give the corrected version and explain what mental model I should be using instead.

Being wrong specifically is more useful than being vaguely right. The model can diagnose your model, not just your words, if you show it what you think.

Make it stick: encoding for later

13. The self-generated analogy.

Don't give me an analogy for [concept]. Instead, ask me to propose one, then critique it: tell me what my analogy captures well, what it distorts, and how to fix the distortion. Give me one hint if I'm stuck.

Self-generated analogies outperform received ones by a wide margin — the generation effect again. Even a bad analogy you made sticks better than a perfect one you were given.

14. Transfer test.

Give me 3 novel scenarios — none from the textbook — where [concept] applies. For each, tell me the surface features that might make me miss the connection, and what deep structure I should be looking for instead. Then quiz me with a 4th scenario where it doesn't apply and ask me why.

You understand a concept when you can recognize it in unfamiliar clothing. Transfer questions train that recognition; definition questions don't.

15. The one-page compression.

I'll explain [concept] as a one-page summary. After I do, grade it: what's accurate, what's misleading, and what's missing that an exam would target. Then help me compress my summary by 30% without losing any critical piece — what do you cut and why?

Compression forces prioritization. You can't compress without deciding what's essential, and that decision is understanding itself. The 30% cut teaches you the shape of the idea — what's core, what's elaboration.

Which prompt when

Not every concept needs 15 prompts. Use this quick selector:

  • First time seeing it: Prompts 1–4 (reframe) → pick the explanation path that resonates.
  • It makes sense but feels fragile: Prompts 5–8 (decompose) → break it into parts you can hold individually.
  • You think you get it but haven't been tested: Prompts 9–12 (test yourself) → retrieval reveals the gap reading hides.
  • You get it now but need it next month: Prompts 13–15 (make it stick) → generate and transfer.

The underlying principle: understanding is not receiving a better explanation — it's producing an explanation that survives testing. AI's role is to give you the test you'd otherwise never encounter until the exam.

Want the structured versions? Use Explain a Hard Topic and Explain at Three Levels for first explanations, Feynman Technique Tutor and Socratic Tutor for retrieval loops — all free in the Promptigo library.