I Tried NotebookLM for Korean. Here's Why I Switched.
I Tried NotebookLM for Korean. Here's Why I Switched.
I was studying for a Korean class and thought I'd found the perfect hack.
I uploaded my textbook chapter, my vocab list, and a grammar PDF into NotebookLM. In about thirty seconds it had summarized everything, answered my questions in plain English, and generated a fifteen-question quiz on the material.
I took the quiz. Got most of them right. Closed the laptop feeling pretty good about my progress.
One week later, I remembered almost none of it.
What NotebookLM Actually Does Well
I want to be fair here. NotebookLM is genuinely impressive for understanding material. If you're confused about why 에 and 에서 work differently, or why a verb conjugation looks the way it does, being able to ask a question and get a clear explanation is useful.
It's also great for building a quick study guide from a dense reading, or for getting a summary of a chapter before you start drilling it.
But there's a gap between understanding something and remembering it. NotebookLM is built for the first thing. It's not built for the second.
The Problem: One Quiz, One Session, Then Gone
When you take a NotebookLM quiz, you're testing yourself once, right after you studied the material. That's about the least effective way to build long-term memory — not because it's wrong, but because it doesn't account for time.
The research on this is clear: the moment you need to review something is not right after you learned it. It's right before you forget it. Miss that window and the memory fades. Hit it at the right time and the memory consolidates.
NotebookLM has no concept of time. Every session starts from scratch. It doesn't know what you got wrong last Thursday. It doesn't know that you've reviewed 에서 six times but still mix it up with 에. It doesn't know anything about you across sessions, because it's a document tool, not a learning system.
What I Actually Needed
After a few weeks of the same cycle — study, quiz, feel good, forget — I started looking for something that would actually track what I didn't know.
The difference, I eventually understood, is spaced repetition. The idea is simple: the app shows you something right before you're about to forget it. Each time you remember it correctly, the next review gets pushed further out. Get it wrong, and the interval resets. Over time, things you know well barely show up. Things you keep forgetting show up constantly.
I also wanted the drilling itself to be more active than multiple choice. Reading a vocab word and clicking the right answer is not the same as producing it from scratch. The format matters.
What I Use Now
I ended up on Chamelingo. A few specific things made the difference:
It remembered what I got wrong. Every word I missed came back. Not randomly, but at a schedule calculated from my actual performance. Within a few weeks, I stopped dreading review because the sessions felt targeted rather than exhausting.
The exercises were varied enough to stay hard. Some sessions I'm typing the Korean from an English prompt. Some sessions I'm listening and writing. Some sessions I'm building the sentence from shuffled pieces. The variety kept it from feeling like I could game the format.
I could bring my own material. My textbook vocab list, my Anki deck, my TOPIK practice words — I imported them and the app generated exercises from them. This was the specific thing I'd tried to replicate with NotebookLM, but here the exercises actually tracked over time.
The multiplayer was a surprise. I started doing 1v1 quiz battles against a friend who was studying Korean at the same time. We set the matches to use the same vocab set we were both studying. Competing against someone you know — and watching them struggle with the same words you struggle with — turns out to be a surprisingly effective motivator. You can also study together in co-op mode (not against each other, just alongside), or even spectate while a friend practices.
NotebookLM Is Still in My Stack — Just Not for Drilling
I didn't delete NotebookLM. It's still useful for the understanding phase: when I encounter a new grammar pattern and want a clear explanation, or when I want to ask questions about a reading.
But for the drilling phase — the part where the language actually gets into long-term memory — I need something that watches what I forget and brings it back at the right time.
That's not what NotebookLM was built for. It was built for document comprehension. It's excellent at that. Language retention is a different problem.
If you're in the same spot I was — using AI tools to study but not seeing the retention stick — it's probably not that you're doing something wrong. It's that the tool you're using wasn't designed to make things stick.
The quiz at the end of a NotebookLM session feels productive. Long-term, it's one-and-done. You need something that comes back.
See the full feature comparison: Chamelingo vs NotebookLM | Why NotebookLM can't make things stick (deep dive)