How to read the labels

Every method page carries a few short labels. They help you choose wisely. They are not a claim that “AI learning is scientifically proven.”

Learning impact is about leverage

Ask: if you use this honestly, how much does it tend to change understanding, skill, or error correction on its own?

Highest-leverage and Strong change are the foundation moves. Lower impact is not a fail. An unlocker can still be the right first step.

Session setup / Unlocks later work , mostly preparation. It clears the path for harder practice.
Clear help / Reliable gain , real help. Worth putting on your schedule.
Strong change / Highest-leverage , big change when you do the cognitive work yourself.

Learning evidence rates the idea, not the chatbot

The evidence label rates the underlying learning idea, retrieval practice, scaffolding, worked examples for beginners, and so on.

It does not mean randomised trials of this exact prompt in this exact app. A strong principle can sit next to an AI delivery that is still lightly tested.

Principle evidence

Is the learning mechanism well supported? Example: retrieving, then getting feedback, beats re-reading the same page again.

AI delivery evidence

Does an AI tutor deliver that mechanism as well as a skilled human, a worksheet, or a good prompt card? Often we do not know yet. Method pages keep this separate on purpose.

Why a “lower impact” method can still be right

Impact is standalone leverage. A strong revisit schedule may depend on quieter setup moves first.

A method that looks modest in general can still be essential in your subject.

What this site is for

Help you tell real learning strategies from convenient shortcuts.

If the AI did the thinking, you practised prompting, not the subject.