How to use the checker, and what it cannot do

The checker on the front page reads a piece of Japanese text and lists the spots that match its rules: common kana and loanword typos, selected adjective and verb endings, doubled particles, bracket pairs and writing consistency. It is meant as a first pass before you, or someone qualified, read the text properly. This page explains how to run it, how to read what comes back, and where its abilities stop.

Running a check, step by step

1. Open the front page and find the text box.
2. Paste your text. The limit is 10,000 characters per check. For a longer document, check it in parts and split at a heading or section break so no sentence is cut in two. Pieces under about 4,000 characters come back faster and give you a list short enough to work through in one sitting.
3. Choose the writing context: everyday modern Japanese, formal or work writing, or literary and historical writing. Literary mode leaves historical spelling and register preferences alone. Quoted wording is skipped by default; enable the quotation option if you want it checked. Then choose the rule groups. For a first pass on a new text, leave all three on: Basic, Readability and Style. The section on settings below explains when to switch one off.
4. Press the check button and wait for the result list.
5. Read the explanation for each finding. When a specific replacement is available, use its button to edit that occurrence in the text box. The check runs again after the edit. Undo restores the previous draft; up to 20 suggestion edits are kept in this tab while it is open. For ambiguous findings, edit the text yourself. Copy current text copies the draft to your clipboard. Keep your master document separately.
6. When you have finished, paste the corrected version and run it again. A second check is quick, and it catches anything a correction introduced, such as a particle left behind after a rewrite.

If you write in a blog editor or a web form, the blog editor page shows a bookmarklet that sends your draft straight to the checker, and the browser shortcut page explains how to send selected text from any page.

Reading the results

Each finding in the list shows four things:

Lines are counted from the line breaks in the text you pasted. A long paragraph with no breaks counts as one line, so for a dense block it is often faster to copy the flagged text and use your editor's search to find it.

If you edit the draft after checking it, the old result list is marked out of date and its replacement buttons are disabled until you check again. Up to 200 findings are displayed per check; use a shorter passage if the limit is reached.

The issue type names match the entries in the legend of issue types, which explains each one with examples and says when a flag is safe to ignore. If you are unsure about a finding, look it up there before changing anything.

Suggestions come in two kinds. Where a rule has a clear answer, the suggestion is a replacement: 一堂に会する for 一同に会する, データ for ﾃﾞｰﾀ. Where it does not, the suggestion is advice instead. A very long sentence gets a hint to split it, not a rewritten sentence, because only you know where the meaning divides. Inconsistent spellings show both forms, and the choice is yours.

Settings: switching rule groups off

The rules are grouped in three levels, and each group has its own switch:

• Basic covers possible spelling slips, selected adjective and verb conjugations, malformed polite endings, suspicious particle/quantity joins, conversion slips, doubled punctuation and unmatched brackets. Keep this group on almost always.
• Readability covers things that slow readers down: very long sentences, too many commas, double negatives and machine-dependent characters. Switch it off for legal or technical text where long sentences are the convention and you are not in a position to change them.
• Style covers optional modern spelling, ra-nuki and redundancy suggestions, consistency and register: mixed desu/masu and da/de aru endings, half-width katakana, mixed full-width and half-width letters and digits, and inconsistent spellings. Switch it off for fiction or interviews full of dialogue, where every quoted line has its own ending and the list would fill with noise.

Web addresses, email addresses and text inside backticks are excluded from wording checks. Paired Japanese or curly quotation marks are also excluded unless you enable checking quoted wording. Brackets themselves are still checked when Basic is enabled. This protects examples and citations from being mistaken for your own prose. ASCII straight quotes are not treated as paired Japanese quotations.

Switching a group off only means its rules are not applied. It never changes your text. You can turn it back on and check again at any time.

What the checker cannot do

The checker combines curated spelling lists, finite adjective and verb rules, punctuation structure and document-level consistency checks. It runs without AI, external APIs or downloaded language models. Its vocabulary is finite; it is not a complete Japanese grammar parser.

• Its grammar patterns cover selected common forms, not every sentence structure. It can flag おいしかた and おいしいでした, but it cannot reliably decide between は and が, resolve an ambiguous modifier or infer a missing subject.
• It does not understand meaning. If the IME gave you 意外 where you meant 以外, both are real words and the checker sees nothing wrong. The same goes for the wrong keigo level for your reader, a date that does not match the source, or a statement that is simply untrue.
• It can flag correct text. Names, dialect, wordplay and fragments may resemble a known typo. Some historical forms are valid in the right context. A style warning is not proof of a grammatical error; leave intentional language unchanged.
• It misses real problems. A clean result list means the text passed these particular rules, nothing more. It does not mean the text is correct.
• It checks Japanese only. English words inside the text are treated as plain characters, apart from the width check.

Privacy

The checker runs in your browser. The rules are part of the page itself, so the text is checked on your own device. Nothing you paste is uploaded or saved by this site. Your draft and the undo history stay in the open tab; the checker does not put them in local storage, cookies or server logs. Copy current text writes only to your device clipboard when you press it. That makes it reasonable to use with drafts you would not want to send anywhere.

Two practical notes. Your text is still in your clipboard and on your screen, so the usual care applies on a shared computer. And the bookmarklet and browser shortcut pass text in the part of the address after the # sign. Browsers do not send that part to the server, but it can remain in your browsing history, so clear the history entry if the text was confidential.

Terms of use, in plain words

• The checker is free to use, for personal or work text.
• It comes with no warranty. It may miss errors and it may flag correct text, and you remain responsible for what you publish or send.
• For anything important, such as contracts, applications, published work or text with legal or financial consequences, have the text checked by a qualified person who reads Japanese well. The checker can make their job shorter. It cannot replace it.
• These terms may be updated on this page.

Linking to the checker

You are welcome to link to the checker from your own site, blog, course page or style guide. No permission is needed. Link to the front page for the tool itself, or to the legend if you want to point readers to the explanations. A plain text link is all it takes:

<a href="https://japaneseproofreader.com/">Japanese Proofreader: a free checker for Japanese text</a>

Please link to the page rather than copying it or showing it inside a frame on another site, so readers always get the current rules.

Questions about using it

Why did it flag something that is correct?

Because it matches patterns, not meaning. The legend entry for each issue type lists the common false alarms. If you know why your text is right, keep it.

Can I check a whole book or thesis?

Yes, in parts of up to 10,000 characters. Check chapter by chapter, keep a note of the spellings you have chosen, and remember that inconsistent spellings are only compared within a single check.

What happens to furigana when I paste?

Ruby text (furigana) usually disappears when copied, or turns into kana in brackets after the kanji. The bracketed kana can trigger extra findings, so remove them before checking if they get in the way.