How the questions are made
JLPT Boi has 8,800 practice questions across N5–N1. They follow the real JLPT's sections and question types, but they aren't official test questions — we write our own.
AI models help draft the questions and explanations. Scripts and review passes check them before they ship, and we keep fixing what learners report. Here's what that means in practice.
Every wrong answer gets explained
When you miss a question, “that option doesn’t fit” tells you nothing. Each wrong option should say what’s actually wrong with it in that sentence. Early versions of our content had lines like the “Generic” ones below. We’ve rewritten them; this is what you’ll see now:
N2 · Word usage
Which sentence uses 見込む (to anticipate) correctly?
Wrong option: 先生を尊敬して見込んだ。
Generic
This option does not fit the meaning or natural collocation of the word in this sentence.
In the app
「先生を尊敬して見込んだ」 misuses it for admiring a teacher; 見込む expects a future outcome, it does not express respect for a person.
Listening audio
The listening audio is made with Google’s text-to-speech voices, not recorded by voice actors. It’s clear and consistent, but flatter than real conversation — the real exam will sound more natural.
Every script is timestamped line by line. The transcript highlights as the audio plays, and in the explanation you can tap a line’s time (▶ 0:14) to hear just that part again.
Other things we check
Every wrong option has its own reason
A check runs over the whole question bank and flags any question where a wrong option has no explanation, shares one with another option, or uses a stock phrase.
Answers are spread across 1–4
Each answer position is correct about a quarter of the time, so always picking 3 won’t get you anywhere. Sentence-rearrange (★) questions never show the pieces already in the right order.
Fixes don’t wipe your progress
When we correct a question it keeps its ID, so your history, stats and review queue carry over.
When something’s wrong
With this many questions, mistakes still get through — a wrong answer key, an unnatural sentence, a typo in a reading. If you spot one, tap the flag on the question screen. Your report comes to us with the question attached, and fixes go out in regular updates.