Privacy
Updated November 6, 2026. This page covers the site jiyomi.app and the Jiyomi app (Android and iPhone).
In short
- The dictionary works without an account, cookies, ads or trackers.
- The app reads your pages on your phone: the photo never leaves the device, unless you ask the AI to reread it.
- We keep only what three things need: the app’s mailing list, AI help, and subscriptions. Nothing is sold.
Who is responsible
Dorian Souc, who publishes Jiyomi. For any question or request: contact@jiyomi.app.
On the site
- What you search for is not recorded anywhere on our side. Like any host, Cloudflare sees your IP address while serving the page.
- The theme and recent words are kept in your browser (local storage), never sent. The ✕ next to recent words clears them.
- Fonts are served by jiyomi.app itself: your browser asks Google for nothing to display them.
- The app’s waiting list (“The app” page): if you leave your address and tick the box, we keep the address, the date and the platforms you ticked (iPhone, Android, tablet, PC, browser extension). It is used to tell you about new versions and about Jiyomi arriving on the stores, and, if you give your Google account address, to invite you to the Play Store test (we then give it to Google Play for that). We keep it until you unsubscribe: every e-mail has a link to do so. To limit bulk sign-ups, we keep a hashed fingerprint of your IP address, for the day only.
In the app
- Reading, furigana and the dictionary run on the phone, offline. Your photos, decks and cards stay on the phone.
- Downloads (dictionary, updates) come from GitHub, which sees your IP address.
- An anonymous account is created the first time the app talks to our server: a random identifier, your plan and the glyphs used this month. The token that protects it is stored only as a fingerprint. To limit abuse, we keep a fingerprint of your IP address on the day of creation, for the day only. We also note the last day the app was opened, and a plain daily count of openings with nothing that identifies you: that is how we know how many people use the app. To count it per platform, for each day the app was opened or synced we keep, for ninety days, the account identifier and the kind of device (Android, iPhone, computer, extension), inferred from the request.
- AI help (translate, explain, reread): when you ask for it, the sentence, with the two sentences before and the two after (or, for a reread, the photo strip of the sentence or the photo of the page), go through our server to Anthropic, which runs the AI (Claude). We keep the answer, so the same request is not paid for twice, a fingerprint of the request and its cost. An image is never kept, only its fingerprint. If another account sends exactly the same request, it may receive the same answer: nobody can look up other people’s requests. When you ask for a translation, the AI also checks how the sentence was split into words and how they are read. If it finds a mistake, we keep the sentence, its words and the correction, without your account, to fix Jiyomi; they are stored in our private space (on GitHub), and erased from the server after thirty days at most. See also Anthropic’s privacy policy.
- Signing in with Apple or Google is optional. We keep only the identifier the provider gives your account: neither your name nor your e-mail address. The first sign-in gives 3 glyphs, once only: so as not to give them again, we keep a hashed fingerprint of that identifier, even after the account is deleted.
- Subscriptions are paid through the App Store or Google Play, or by card on a Stripe page from the PC or the extension: we never see your card. RevenueCat, which tracks subscriptions for us, receives your account identifier and the subscription status; we keep the plan and its end date. For a card payment, Stripe, the seller, receives your card, email address and country, under its own responsibility.
- Reports and ideas you send us (Settings → Report · suggest, or “Report” on a word): your text, the sentence or word concerned and how it was split, the app version, the phone type, the language you read and the state of the dictionary. They go through our server, then are stored in a private space (on GitHub) to be read, and erased from the server. Nothing leaves without that tap; a report that could not be sent waits on your phone and leaves on its own later. On the computer and in the extension it is the same (Settings → Report · suggest, “Report this reading”, or “Report” on a word card): your text, the version and the system (for example “linux KDE” or “Brave 140”). The screen capture and what Jiyomi read in it only leave if you tick the box, unticked by default; so does the address of the web page. To limit abuse, we count what is sent per account and per day, under a fingerprint of the IP address kept for the day only, and we keep a fingerprint of each text received for forty days (the fingerprint, not the text), so the same one is not stored twice. A short summary of each report (its date, platform, version, the word reported and the first 500 characters of the text, without your account) stays on our server for one hundred and eighty days, so we can count reports; it is erased after that.
- Getting an e-mail when a report is sorted (optional, a box unticked by default under the form): if you tick it, we keep your e-mail address and the app’s language next to that report, separate from it (it never goes to the GitHub space) and without your account. It is only used to write to you twice at most: when the report arrives, then when it is fixed or set aside. It is deleted after that last e-mail, or after one hundred and eighty days at the latest. The first e-mail has a “Stop these e-mails” link that deletes it straight away. The e-mails are sent through Cloudflare, which already hosts the server. Your address is never used for anything else or added to any list.
- Your last 5 photos, only if you agree: the app asks once, box unticked by default. They stay on your phone and only leave (with what the app read in them) if you tick “attach my last pages” when sending a report. Settings → “Keep my last 5 photos”: OFF erases them.
- Sync across your devices (free with a Google or Apple account; card photos with a subscription): when it is on, your decks, your cards (the sentence, the kept word, the translation if you asked for it, the printed furigana), your reviews, your notebook notes, the titles of your books and the page you are at, and your companions’ experience go through our server as a log (table
operation) to reach your other devices. Your cards’ photo strips are kept in a separate file store (Cloudflare R2), in the European Union, under your account identifier. A book’s file (PDF, EPUB) never leaves the device: only your position does. None of this is read by us or anyone: your account reads it back on the other device. It all goes with the account when you delete it. - Linking a device (Settings → Sync): an eight-character code, kept as a fingerprint for ten minutes, gives the other device a token for your account. For each linked device we keep the name it gave itself, the date, and the date of its last sync: that is the list you see, and “Unlink” removes it. To limit code guessing, an IP fingerprint is kept for the day.
- Deleting your account: Settings → Account → Delete my account. Your account, sign-ins, linked devices, sync log, photo strips, kept answers and reports not yet collected are erased. Costs already paid stay in our totals, with no link to you any more. Without the app (uninstalled, phone lost): write to contact@jiyomi.app; we check with you that the account is yours, then erase it, with everything above, within 30 days.
The browser extension (coming soon)
- What goes to jiyomi.app: to find the words of a page, the extension splits it inside your browser and only sends short pieces of text to look up in the dictionary, never the page’s address, never the whole page. The server keeps none of it (no log, no cookie); as for the site, Cloudflare sees your IP address while it answers.
- What it keeps: your settings, your cards and decks, and the answers already received, in the extension’s storage, on your computer.
- Video subtitles: on YouTube, the extension reads the subtitles on screen and the subtitle file the player already downloaded, inside your browser; nothing from the video goes to jiyomi.app. Once a day, it fetches from jiyomi.app a small file that says where to find the subtitles in the page. If it can no longer find them while they are on, it sends a signal that only says the site, the reason, the extension’s version and that file’s version: not the video, not the page, not your account. At most one per day and per reason, kept for ninety days as a counter. This signal does not exist in Firefox. When you keep a sentence during a video, its title and link stay in the extension, on your computer.
- Wikipedia: when you open the card of a name (a person, a place, a work), the extension asks Wikipedia whether an article exists, sending that name and a few characters around it in the sentence, never the page or its address. Wikipedia (Wikimedia Foundation, United States) then sees your IP address, as when you read an article.
- If you link it to your account (code shown by the app): your cards go through synchronisation, exactly as described above for the app.
Who else processes this data
- Cloudflare (USA): hosts the site, the server and its database, and sends the report follow-up e-mails; synced cards’ photo strips live in its European Union file store.
- Anthropic (USA): the AI behind AI help, when you ask for it.
- Google: Google sign-in and Google Play in the app.
- Apple: Apple sign-in and the App Store.
- RevenueCat (USA): subscription tracking.
- Stripe (Ireland, USA): card payment, as the seller of the subscription.
- GitHub (USA): downloads, and the private space where reports are stored.
- Wikipedia (Wikimedia Foundation, USA): when you open a name’s entry, your device asks Wikipedia for its article (the name and a few characters around it in the sentence, never the whole page). We keep nothing of that request.
When data leaves the European Union, it does so under a framework provided by the GDPR (Data Privacy Framework or standard contractual clauses).
Legal basis
The mailing list, reports and kept photos rest on your consent (the box, the button), which you can withdraw at any time. The account, AI help and subscriptions rest on what is needed to provide the service you ask for. IP fingerprints rest on our legitimate interest in preventing abuse.
Your rights
You can ask to see, correct, take away or erase your data, and object to its use: write to contact@jiyomi.app. You can also complain to your data protection authority (in France, the CNIL).
Minors
The mailing list, sign-in and subscriptions are for people aged 15 and over, or with a parent’s consent.