Methods · Language & communication · Short guide
G2
Graded reader generation
Work through text rewritten to a controlled vocabulary: a ladder of versions that build toward the authentic source.
When to use it
- You have a real article or story chapter you want to read, but the first page knocks you out.
- You want the same plot or argument several times, each time with a slightly richer vocabulary.
- You keep skipping straight to “authentic” text and never finishing it.
What to ask AI
Type something like: “Target text: [paste]. Make graded version 1 for B1 with controlled vocabulary; list new items (max 12). I’ll retell before version 2. Do not jump to the original until I ask.”
Ask for a ladder of versions toward one authentic source, not a random new story each time.
Keep the hard part on your side: you retell each rung before the next expands.
How it helps
One hard authentic page is often too far a jump from where you are.
A graded reader path lets you meet the same story or argument again and again while the vocabulary expands.
That repetition builds reading stamina in a way a single incomprehensible pass does not.
This is different from one-shot comprehensible input: here the destination is a specific target text, and the rungs are deliberate steps toward it.
How to run it
- Pick a short target text, a news paragraph, a story page, a course reading.
- Ask for version A at your current level, with a capped list of new items.
- Read and retell before you request version B.
- Climb to a richer version C, still holding the same plot or claims.
- Finish on the original. Note only what still blocks meaning.
- Keep a vocab diary only for words that recur across rungs, not every novelty.
Example exchange
I've got a French article about city bikes. Can you rewrite just the first two paragraphs for A2, with at most ten new items? I want to retell before you show me anything closer to the original.
Version 1 is ready. Ten new items are listed at the end, ignore them until after your retell. What happens in the text?
The city added more bike stations. Some people complain about traffic. The mayor says it reduces pollution.
Solid. Version 2 will put back three of the original nouns you skipped. Retell again after you read it, we only open the authentic paragraphs when you ask.
Copyable prompt
Target text: [paste or describe].
Make graded version 1 for [LEVEL] with controlled vocabulary; list new items
(max 12). I'll retell before version 2. Do not jump to the original until I ask.
The Tell
Here is how you know this method has flipped: you stay forever on comfortable graded versions and never open the authentic text.
The ladder exists to reach the real page, not to replace it.
Schedule the original as a finish line, and treat endless “one more easy version” as avoidance.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
The underlying learning idea is rated moderate. Graded readers sit in a long classroom tradition of extensive reading with controlled vocabulary. The claim is practical: volume of mostly understandable text builds fluency. How tightly AI can control vocabulary while keeping the same story faithful is a separate, less settled question.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
There is no strong evidence that AI-made graded ladders match carefully edited human graded readers. Treat fidelity to the original claims and true vocabulary control as speculative; check the final authentic text yourself.