Every roundup of Japanese grammar apps answers the same question. Which one has the most grammar points, the clearest explanations, the smartest review algorithm. Those are fair questions and I will answer them below, with prices and a comparison you can act on.
But there is a question that comes before all of them, and almost nobody in English asks it.
Japanese has two grammars.
Not two opinions about grammar. Two complete descriptions of the same language, both taught in Japan today, both correct, and different enough that they disagree about what the parts of a verb even are. Japanese children learn one of them in middle school. Every grammar app aimed at you teaches the other one. Very few products mention that a choice was made, and the gap between the two quietly explains several things learners usually blame on themselves — including why asking a Japanese friend a grammar question so often goes nowhere.
I am Japanese, I live in Tokyo, and I sat through six years of the grammar your app does not teach you.
What I earn from this page
Updated 22 August 2026, because I said on this page that I would say so when it changed. JapanesePod101 has approved me, at 25 percent, on a cookie that does not expire. The Grammar Bank section below carries a tracked link. Updated again 4 September 2026: I now hold four approved affiliate partnerships. Two of the others are Japanese knife retailers and have nothing to do with this page. The fourth is The Japan Shop, a bookshop approved on 4 September 2026 which does sell Japanese grammar books, so that one is not unrelated to this page, and there is no link to it on this page. Other pages here do carry links to that shop, the largest being in my guide to starting to read Japanese, and my affiliate disclosure page keeps the full list current. Rocket, which runs the highest paying programme of anything here, is not one of them, so it pays me nothing. Nothing else on this page pays me anything, and I have not moved a single product up or down since the approval came through.
I am telling you this because of where the recommendations land. The product I put first has no cash referral program at all. The product that would pay me the most, at forty percent, is one I have to rule out as your main grammar resource for a reason I will show you. That has now happened in every article in this series, and it is not a pose — it is what falls out when you check the products against what the language actually requires.
Japan has two grammars of Japanese, and they do not match
The National Institute for Japanese Language and Linguistics runs a public question and answer service. One of the questions is exactly the one at the top of this page: is the grammar foreigners learn the same as the grammar taught in Japanese middle schools and high schools?
The answer is written by Inoue Masaru, a professor in the College of Humanities and Sciences at Nihon University. His conclusion, in the first line: as things currently stand, they are considerably different. The piece first appeared in a National Institute publication in 2002 and was reissued on the institute site in 2023.
He gives two reasons, and both matter to you.
Reason one: there is more than one way to cut a verb
Take the verb 書く, to write. Where does the stem end and the ending begin?
School grammar, the one Japanese people learn, says the unchanging part is か and the endings are く, け, こ. So 書く is か plus く, 書け is か plus け, and 書こう is the irrealis form 書こ plus the auxiliary う.
Japanese language teaching grammar, the one your app uses, says the stem is kak- and the endings are -u, -e, -oo.
Both descriptions are true. They cut in different places. And the reason they cut differently is worth sitting with, because it is not arbitrary.
School grammar cannot cut inside a kana. か is one character and one mora, and the writing system offers nothing smaller. So the smallest piece school grammar can name is か. The romanised analysis splits the kana く into a consonant k and a vowel u, and once you allow that, Japanese conjugation becomes beautifully regular — one stem, a tidy set of endings, no exceptions to speak of in the largest verb class.
That is a real advantage, and it is why the teaching grammar chose it. But notice what you lose. In the school analysis, the endings く け こ か き are simply the か row of the kana grid, read straight down. Conjugation is a movement across the writing system. If you learned kana from a grid rather than a list, you already own the machine that generates 書か, 書き, 書く, 書け, 書こ, and nobody told you, because the analysis you were handed does not use it.
Two descriptions, both correct, predicting different things. Your app gives you one and lets you assume it is the only one.
Reason two: school grammar covers only part of the language
This is the heavier point, and it is the one that costs you sentences.
Inoue gives an example. Compare these two:
- 太郎は体が大きすぎて茶室に入れない。 Taro is too big to fit into the tea room. Correct.
- ✕ このピアノは大きすぎて部屋に入れない。 This piano is too big to go into the room. Wrong. It has to be 入らない or 入れられない.
Same verb. Same potential ending. One sentence is Japanese and the other is not.
The rule is that the potential form can be built only when the verb expresses a deliberate action by someone. When the verb expresses nothing more than a change brought about by something else, as with a piano being too large for a doorway, the potential form is unavailable. Taro chooses to enter. The piano does not.
Here is Inoue on what school grammar does with this: for native speakers it is normal to just say that 入れる expresses potential and leave it there. For Japanese language teaching, he writes, knowing in which cases a verb can take the potential form is extremely important information, because without it you cannot use the form appropriately.
So the constraint is not hiding in the native grammar where learners cannot reach it. It is in neither description. School grammar skips it because native speakers never need it stated. And that, I think, is why no product has picked it up.
I went looking for a product that teaches the constraint
This series has a habit of assuming no product covers something and then going to check, because twice now the assumption has been wrong and the correction made the argument stronger. So I checked.
I went to the most careful grammar product on the market. Bunpro has more than nine hundred grammar items, each with a write up, and its entry for the potential form sits at N4.
That entry is genuinely good. It covers how to build the form for both verb classes. It covers the する and くる exceptions and the fact that する borrows できる instead. It covers ら抜き言葉, the dropped-ら colloquial forms that Japanese people argue about. It covers the ambiguity where 食べられる can mean either can eat or is eaten, and how you tell from context. It covers when to use 見える and 聞こえる instead.
It does not mention the constraint. There is no line saying the potential attaches to deliberate actions only, and no example along the lines of 入る or 開く.
I want to be careful about what that does and does not prove. It proves one entry in one product, checked on one day. It does not prove that every app everywhere is silent. But the entry I checked is the most thorough treatment of the potential form I can find in an English-language product, from the app most likely to have caught it, and the omission is not an oversight by a sloppy team. The information is missing from the source material both grammars are built on.
Which gives the shape of the problem in this whole category:
A grammar point, as shipped by every product I looked at, is a form plus a meaning. Production needs a form, a meaning, and the conditions under which the form may be used. The third part is the part that is missing, and it is the part you cannot reconstruct from examples.
Readers of the rest of this series will recognise that. It is the same hole as pitch accent taught but never tested, politeness levels taught but the choice between them never posed as a question, counter lists indexed in the direction you will never need. Fourteen articles in, it keeps arriving in the same form.
Why this breaks the advice to just ask a native speaker
Every guide tells you to practise with Japanese people. It is good advice. But it fails in one specific place, and grammar is that place.
Your app taught you te form, Group 2 verbs, na-adjectives, dictionary form, i-adjectives, plain form. A Japanese adult has never encountered a single one of those terms. They are not simplified versions of what we learned. They belong to a different system.
What we learned instead was 連用形 and 連体形 and 形容動詞 and 助動詞 and 文節, a ten-part-of-speech classification, and a six-form conjugation paradigm. That system was assembled under the linguist Hashimoto Shinkichi for a Ministry of Education textbook in 1943, rewritten for the postwar curriculum in 1947, and — this is the part that surprises people — it deliberately includes classical Japanese grammar alongside the modern language, because in Japanese schools grammar exists largely to prepare you to read classical texts. Japanese commentary on it notes that it has come down to the present without major revision since the 1950s.
So when you ask a Japanese friend how to make the te form, three things happen at once. They have no idea what a te form is. They can produce every te form in the language flawlessly. And the grammar they were taught in school would describe 書いて as the continuative form plus a conjunctive particle, with a sound change applied, which is not going to help you at a bar.
This is the structural explanation for something I raised in the first article in this series: native speakers produce the language perfectly and cannot explain it. That is not modesty or bad memory. Half of what you want to ask about was never in our description, and the half that was is filed under labels you have never seen.
What we call the things your app taught you
If you do want to talk grammar with a Japanese person, here is the conversion. The middle column is what an ordinary adult might recognise from school. The right column is what I actually recommend saying.
| Your app says | We were taught | Better to say |
|---|---|---|
| te form | 連用形 plus て, with sound change | Just say the word: 「書いて、みたいな形」 |
| dictionary form | 終止形 | 「辞書に載ってる形」 |
| Group 1 verb / u-verb | 五段動詞 | Give two examples instead |
| Group 2 verb / ru-verb | 上一段・下一段動詞 | Give two examples instead |
| Group 3 / irregular | サ変・カ変 | 「する と くる」 |
| na-adjective | 形容動詞 | 「きれい、みたいなやつ」 |
| i-adjective | 形容詞 | 「高い、みたいなやつ」 |
| particle | 助詞 | 助詞 works, most adults know it |
| helper verb, ending | 助動詞 | Give the example |
The honest advice underneath that chart is: do not use metalanguage with native speakers at all. You are asking a fluent speaker to do linguistics, which is a separate skill they do not have and did not sign up for. Ask about sentences instead.
Three questions that work
These three get useful answers from any Japanese person, with no shared vocabulary required. I have been on the receiving end of all three and they are easy to answer well.
- 「この言い方、変ですか」 — Is this way of saying it strange? This is the highest-value question in the language, because natives are extremely reliable detectors of strangeness and extremely unreliable explainers of it. Take the yes or no and do not push for the reason.
- 「AとB、どっちが自然ですか」 — Which of these two is more natural? A forced choice between two sentences you wrote yourself. This is the only way I know to get at the conditions of use that your app left out, and it is exactly how you would have caught the piano sentence.
- 「これ、どういう時に使いますか」 — When would you use this? Note that it asks for situations, not rules. Native speakers are good at situations. Ask for a rule and you get a wrong rule delivered with confidence.
If you have nobody to ask, an AI voice partner will answer these three all day, with the important caveat set out in that article: it will correct your grammar and stay silent about everything social.
Why a grammar point is the unit in the first place
Once you notice that products sell grammar as a numbered list of points, the next question is where the list comes from. It is not from either of the two grammars above. Neither of them is organised that way.
It comes from the test.
Look at how the libraries are arranged. The Bunpro grammar library is grouped as N5, N4, N3 and so on, subdivided into numbered lessons. The JapanesePod101 Grammar Bank is a database you filter by JLPT level, category and series. The PORO grammar app advertises more than seven hundred grammar points covering almost all JLPT levels, with more than a thousand questions imitating JLPT question types.
The grammar point is a JLPT item. That is the unit, and every product inherited it.
I have written about what the JLPT does and does not measure, and it is the hub of everything in this series: the exam has no speaking section and no writing section at any level, and every question is multiple choice. So the unit of grammar that the entire market is built on is scoped to what a multiple choice question can ask. A form, a meaning, four options.
There is a version of this argument I do not want to make. When I looked at counter apps I argued that the item count was the wrong axis, because you only need a couple of dozen counters and products were selling eighty four. Grammar is not like that. You really do need most of nine hundred grammar points eventually, and Bunpro is not padding. The count is honest.
The problem is not the size of the list. It is that each item on it is shaped like a test question, so the part of each item that a test question cannot ask about never got written. The piano sentence is not a thing a multiple choice item can probe. So it is not in the point. So it is not in the app.
The three camps, and why nobody tells you which one they are in
Here is the practical payoff of everything above. English-language grammar resources fall into three camps, and mixing them in your first year is the single most avoidable source of confusion in this category.
Camp one: teaching grammar. Bunpro, LingoDeer, the JapanesePod101 Grammar Bank, Rocket Japanese, PORO, and effectively every app you will meet. Groups one, two and three. Dictionary form, ます form, て form, ない form. i-adjectives and na-adjectives. This camp is the default and, for a beginner heading for production, I think it is the right default.
Camp two: its own analysis. The Tae Kim guide is the notable one, and it argues its choice out loud, which no app does. On why textbooks start with the polite form, Tae Kim writes that they want readers to be able to use functional and polite Japanese as quickly as possible, and that they probably decided to use the polite form even though learning the polite form before the dictionary form makes no sense. His stated principle is to take Japanese and explain how it works, and forget about trying to force what you want to say in English into Japanese. Whether or not you accept it, you are told what you are buying.
Camp three: traditional terminology. Imabi is the one substantial English resource here. It uses 未然形, 連用形, 終止形, 連体形, 已然形 and 命令形, names verb classes as 四段, 上一段, 下一段, サ変 and カ変, and teaches classical Japanese as part of the curriculum rather than as a curiosity, including how the nine classical verb classes collapsed into the five modern ones. It is the closest thing in English to what I was taught.
The blogger behind nihonshock laid the two systems side by side years ago and summed the trade off up fairly: the simplified system sufficiently accounts for most needs of beginners in understanding how Japanese conjugates, but for those hungry for minute details of grammar, you will eventually run into roadblocks with this method.
Nobody labels their camp on the sales page. So a motivated learner ends up running Bunpro for reviews, reading Imabi for depth, and asking a tutor for clarification, and receives three different names for one verb ending, in three metalanguages, and concludes that Japanese grammar is chaotic. It is not. The description is fragmented, and the fragmentation is invisible.
Pick one camp for your first year. Camp one, almost certainly. Visit camp three later, deliberately, when you want to read older texts or talk to Japanese people about their own language.
The comparison
Prices are what the official pages showed when I checked in August 2026. The final column is the disclosure that matters most in this category.
| Resource | Grammar system | What it makes you do | Price | Can it pay me? |
|---|---|---|---|---|
| Bunpro | Camp one, JLPT ordered | Type the answer into a gap | Free tier forever, $5 per month, $150 lifetime, 30 day trial with no card | Nothing. Referrals earn cosmetic points only |
| Tae Kim guide | Camp two | Read, then self test | Free | Nothing |
| Imabi | Camp three | Read | Free | Nothing |
| Tofugu grammar reference | Camp one | Look things up | Free | Nothing, and it is a competitor |
| JapanesePod101 Grammar Bank | Camp one, JLPT filtered | Listen and read explanations | Premium feature, 7 day free trial | Yes, 25 percent |
| Rocket Japanese | Camp one | Listen, repeat, some recording | Paid course, caps at N4 | Yes, 40 percent, the highest here — I am not in it |
| LingoDeer | Camp one | Tap and arrange | Not stated on the official page | Could not verify a program |
| PORO Japanese Grammar | Camp one, JLPT shaped | Multiple choice quizzes | Free with ads, store listing only | Unknown |
| A Dictionary of Basic Japanese Grammar | Camp one, reference | Look things up | Print book | Nothing, I am not in any book programme |
What I would actually use
Bunpro, as the spine. It is the only product in this category that makes you produce the form rather than recognise it. You type into a gap, it marks you, and the review scheduling is genuinely good. Nine hundred plus items, ten thousand plus example sentences with native audio, a hundred and twenty graded reading passages, and a link to WaniKani if you use it. The free tier is permanent and the thirty day trial needs no card, so you can find out whether typing answers suits you before spending anything. At five dollars a month it is the cheapest serious thing in this whole series. Its referral programme hands out points for profile cosmetics, so recommending it earns me precisely nothing, which is the fifteenth time in fifteen articles that the top pick has been unmonetised.
The Tae Kim guide, read once, early, in full. Free, opinionated, and short enough to finish. Read it for the argument, not as a reference.
Imabi, later and deliberately. Free. Go there in your second year, or immediately if you plan to read anything written before 1945, or if you want to be able to discuss Japanese grammar with Japanese people in their own terms.
The Tofugu grammar reference, free and good. It competes with this site and I am recommending it anyway, same as with their counter list.
JapanesePod101 Grammar Bank — and here is my conflict of interest. This is the product on the page that pays the most per sale after Rocket, at twenty five percent, and as of 22 August 2026 I am approved for that programme, so the link above is a tracked one. So take the following with that in mind. The Grammar Bank is a well built database: filter by JLPT level, category and series, with romanised and kana examples, audio and written explanations on every item. If your bottleneck is that written explanations do not stick and you need to hear the sentence, it is a real solution and the seven day premium trial costs nothing to try. Past the trial, every paid subscription carries a 60-day money-back guarantee, stated in the FAQ at the foot of their pricing page and checked again on 27 August 2026; you claim it by email rather than by a button, and a refund is deducted back off my commission. But it is a bank of explanations. Nothing in it requires you to generate a form under any pressure, which is the same limitation I found in its business Japanese material. Use it beside Bunpro, not instead of it.
Rocket Japanese — the highest paying product here, and I cannot recommend it for this job. The Grammar and Culture series runs thirty one lessons paired to thirty one audio lessons, and the audio course is a good product for other purposes. But Rocket states on its own site that Level 1 corresponds to N5 and Levels 2 and 3 to N4. A grammar spine that stops at N4 cannot be your grammar spine. It is the fifth article running in which the forty percent product has come up short on the specific thing the article is about, and I am not going to start pretending otherwise now.
LingoDeer — I could not verify it. Its grammar-led curriculum is well regarded and it was built for East Asian languages rather than retrofitted, which usually shows. But the official site does not state pricing on the landing page and its partnership section covers business and educator programmes only, with no affiliate programme I could confirm. I am not going to quote a price or a commission I cannot source, so it stays in the comparison above unrated.
A Dictionary of Basic Japanese Grammar, the yellow book. Still the best explanation of any single grammar point you are stuck on, and I have no way to earn from books, so treat this as a straight recommendation.
What none of them do
I will end where the series always ends up.
Every product above will teach you the potential form. One of them will make you type it. None of them will tell you that the rule they just taught you produces a wrong sentence when the subject is a piano. That information sits in a free question and answer page on the website of a Japanese national research institute, written in Japanese by a professor, and it has been publicly available since 2002.
This is the same shape as the keigo classification the Agency for Cultural Affairs updated in 2007 and no English product noticed, and the writing-style rule Japanese nine year olds learn under the national curriculum that no English product teaches. Japan documents its own language carefully and publishes the documentation. The documentation is in Japanese, so the market that serves English speakers does not read it.
So the practical instruction is smaller than it sounds. Buy a grammar app — Bunpro, most likely. Work through it. But hold the rules loosely, because each one arrived at you with its conditions of use stripped off, and the only way to put them back is to write your own sentences and get a human being to tell you which ones are strange. That is a five second question in Japanese, and now you know exactly how to ask it.