Read Japanese (iOS / iPadOS / macOS) — last updated: 27 August 2026
Read Japanese (the “app”) respects your privacy. The app does not require an account, and we never ask you for your name, email address, phone number or anything else that identifies you. The one exception is the feedback box (section 3): it sends exactly what you type, so anything you write there reaches us.
The works you have read, your reading position, bookmarks, highlights and notes, the words you looked up and their history, and your reading log are stored on your device. The developer never receives them automatically. They reach the developer only if you type them into the feedback box yourself (section 3).
If you turn on iCloud sync (a Pro feature), these are also stored in the iCloud private database of your Apple Account and carried between devices signed in to the same Apple Account. That area is managed by Apple, and the developer cannot read its contents. Without Pro, and with sync turned off, everything stays on the device alone. You can turn sync on or off at any time in Settings > Sync in the app.
The text of a downloaded work is stored per device, whether sync is on or off.
To improve the app we use Google’s Firebase Analytics, Firebase Crashlytics and Firebase Performance Monitoring to collect the following.
| What we collect | Why |
|---|---|
| Which features are used (a work was opened, a word was looked up, read aloud was used, and so on) | To find features people never reach |
| The work ID of an opened work (the number 青空文庫 (Aozora Bunko) publishes for each work) | To find works where the layout breaks |
| How far you read, as a stage (first page / 10% / 50% / finished), and the stage where you stopped | To learn where people stop reading |
| The category of the work (top level of the Nippon Decimal Classification) and the length of the text, as a stage | To learn what kinds of works are read |
| Reading time per session, as a stage | As a guide to readability |
| That a purchase or a restore happened, and which kind (monthly / yearly / lifetime) | To learn which form of pricing is chosen (section 6) |
| How long parsing and pagination took | To find slow places |
| Crash records (where it happened, device model and OS version, the work ID and page number open at the time, type size and writing direction) | To fix bugs |
| Annotation syntax from 青空文庫 that we failed to interpret, and the work ID | To find notation we do not yet support |
| An identifier for the device (an app instance ID for analytics, and, if you have turned notifications on, the token that addresses them) | For the above, and to deliver notices (section 5-2) |
| The approximate region inferred from your connection (country or region level) | For the above. We do not ask for location permission and never read the device’s location |
What we do not send: titles, author names, the text of works, the words you looked up, and the contents of your notes on bookmarks and highlights are never transmitted.
About work IDs. A work ID is the number 青空文庫 publishes for each work. It is public information that belongs to no one. We collect no name or email address, and we do not connect a work ID to any individual.
Why crash records include a page number. The app performs Japanese vertical typesetting itself (furigana, emphasis dots, tate-chu-yoko, inline notes, gaiji). Faults occur on one particular page of one particular work, and cannot be fixed without knowing which. The page number is a count of pages, not the content of the text.
What we collect is treated statistically and is never used to identify an individual. You can stop analytics and crash reporting separately in Settings > Privacy in the app.
What you send from Settings > Feedback > Send feedback goes to a Google Cloud Firestore database run by the developer, and the developer reads it. We receive the following.
| What is sent | Why |
|---|---|
| Exactly the text you typed | To read what you told us |
| App version and build number, OS version, a string identifying the device model (such as iPhone17,1), and your language and region settings | To isolate faults that affect one version, model or language |
| The time it was sent | To read them in order |
The text is sent as written. If you type your name, email address or phone number, that is sent too. Please do not write anything you would rather not share. For anything that needs a reply, please use the contact address in section 11.
Nothing is sent unless you press Send. No account is needed, and we attach nothing that says who you are (no account, no advertising ID, no unique device number). Only the developer reads what you send; it is not passed to third parties (section 7).
Your reading data (the text, reading position, bookmarks, highlights and notes, the words you looked up) is never sent unless you type it into this box yourself.
The app asks for microphone and speech recognition permission only when you use Read-Along. It matches what you read aloud against the text to follow where you are. Audio is processed on the device only — nothing is recorded, stored or transmitted. On macOS you can revoke or restore this in System Settings > Privacy & Security > Microphone.
On Mac there are two routes to a definition.
The app connects to the network in the following cases.
The AI that explains words runs entirely on your device through Apple’s Foundation Models. No text is ever sent to an external generative AI service. Word translation also uses on-device translation and does not use the network.
We use two mechanisms to tell you about new features and newly added works. Both can be turned on and off in Settings > Notices in the app, and both are off to begin with. Nothing below is sent until you turn one on.
Notices (push notifications)
Only when you turn “Receive notices” on does the app ask for notification permission and register this device with Apple’s APNs and Google’s Firebase Cloud Messaging.
| What is sent | Why |
|---|---|
| Identifiers that address delivery (the APNs device token, and a registration token issued by Google) | To deliver notices to this device. Held by Google |
| Membership of a group that receives the same notice | To send one notice to everyone rather than to individuals |
These identifiers address a device and are not connected to a name or email address. Every notice goes to everyone alike; we do not vary them by what you have read or by your reading log.
How to stop them. Turning “Receive notices” off in Settings > Notices removes the group registration and discards the registration token held on the device. You can also stop them from the notification settings of the Settings app (System Settings on macOS).
In-app messages
Only while “Show in-app messages” is on does the app display messages delivered by Google’s Firebase In-App Messaging. Whether one was shown, tapped or dismissed is included in the analytics in section 2. Nothing interrupts you while you are reading.
The app offers a paid plan covering some of its features (reading the works themselves is free).
For stopping auto-renewal and for refunds, see the support page (section 11).
Except where required by law, we never provide or sell what we collect to third parties. The app shows no advertising, and there is no third-party advertising tracking.
The app carries works of modern Japanese literature (from the two sources in section 9). Some works contain adult expression, and the look-up feature can open any web page, so the app is intended for the age range set by its App Store age rating.
Every work in the app is one whose copyright term has expired in Japan. There are two sources, with different grounds and different delivery. The work detail screen shows which source a work came from.
青空文庫 (Aozora Bunko) — 17,344 works
Following the Aozora Bunko guidelines for handling its files, each work shows the source edition, the person who input the text and the proofreader, in addition to the title and the author. Information derived from the library cards is used under Creative Commons Attribution 4.0 International.
National Diet Library Digital Collections — 28 works
These 28 works, mostly translations, come from materials marked with the Public Domain Mark (PDM) in the National Diet Library Digital Collections. They were not input and proofread by volunteers as Aozora Bunko works are, so they have no transcriber or proofreader. Instead, the work detail screen shows the bibliography of the material it was made from and a link to the material itself.
The text was produced by optical character recognition (OCR) of the material and then tidied by us for readability. PDM materials may be used in modified form, but the modification must be stated, so we state it at the end of the text and on the work detail screen. Errors left by OCR may remain.
Before including a work we checked, one by one, the year of death of the author and also of the translator and of anyone who wrote a commentary, and carried only those whose term has expired in Japan.
The text of these 28 works is fetched from a separate host that we run (novel-b5536-packs.web.app). The app verifies with an embedded checksum that what arrives has not been altered in transit.
If we change this policy we will update this page. For significant changes we will also say so in the release notes.
Please send questions and requests to r.muranaka.work@gmail.com.
This policy covers the iPhone, iPad and Mac app “Read Japanese” — see the support page.