Last updated: 27 October 2026
Volex is a vocabulary-learning app. This policy explains exactly what leaves your device, why, and who else can see it. Volex shows no ads, contains no third-party analytics or tracking SDKs, and never sells or shares personal data. It keeps its own daily usage counts, which you can turn off — see “Usage statistics” below.
Like the rest of your iPhone’s app data, everything in this list is also included in your own device backups — iCloud Backup or a backup on your computer — if you make them. With Volex Pro, iCloud Sync also keeps your vocabulary and study progress in your own iCloud (see “iCloud sync” below). We cannot access these copies.
| What | Where it goes | Why |
|---|---|---|
| The word you are adding, plus the language pair and the fields you asked for | Our Cloudflare Worker, which forwards it to Google Gemini | To generate the definition, examples, synonyms and translations you requested |
| Text you ask to translate | Google Cloud Translation; English dictionary definitions come from dictionaryapi.dev | To return the translation or definition |
| While translation suggestions are on (Settings → Translation suggestions), the word you type when adding a new word, plus the language pair | Our Cloudflare Worker. It answers from our reviewed word list; only when the word is not there, it asks Google Cloud Translation or Microsoft Azure Translator | To suggest translations. The word is kept in a shared translation cache and in anonymous counts of looked-up and chosen words — never together with your device identifier. The device identifier is used only to count your free suggestions, to limit requests per device and to check an organisation licence |
| The word whose pronunciation you play with a speaker button or that a Listening quiz question plays, and in Listen mode the text the session speaks — the word, its translation and the example sentence | Our Cloudflare Worker, which forwards the text to Google Cloud Text-to-Speech, or to Microsoft Azure Speech for Uzbek and Kazakh | To produce the spoken audio you hear. Nothing is sent for a word that already has your own recording or a downloaded pronunciation file. Audio already received is kept on your iPhone and played from there, and your iPhone’s built-in voice takes over whenever the cloud has no voice for the language or does not answer in time |
| A photo of a page you choose to have read by AI (Volex Pro or the notebook promo) — a vocabulary list, or a book page with the words you marked: resized, with location data removed on your phone | Our Cloudflare Worker, which forwards it to Google Gemini | To read the words on the page — or only the words you marked, with their sentence and a translation — into a list you check before anything is saved. Our server keeps neither the photo nor the words read from it; under Google’s paid API terms Google does not use it to train its models |
| A random device identifier | Our Cloudflare Worker | To apply the daily free limit and stop abuse, and, with a random marker created when the app is installed, to give the notebook promo once per device. It is a random value created on first launch, is not your Apple ID, not an advertising identifier, and is not linked to your name or email, unless your organisation entered a name or label for an access code you redeemed |
| Feedback you send: your message and, if you add one, a contact (an email address or Telegram username). When you report a mistake in a word, also the word and the content you see for it. When you tap “Send correction” after editing a word, the word and the previous and new values of only the fields you changed. When you report a problem with spoken audio, the line’s text and language are sent; we keep only an identifier of the recording and a one-way fingerprint (hash) of the text, not the text itself. If we replace that recording, the fingerprint and language are added to a public list the apps read to delete their saved copy. Each carries your app version, platform and interface language, and — if you use an organisation access code — your organisation’s identifier | Our Cloudflare Worker | To read and answer it, and to correct AI content for everyone. Nothing is sent until you tap Send or “Send correction”; a correction is never sent on its own |
| Usage statistics (can be turned off in Settings → Privacy): a random identifier created for statistics only, separate from the device identifier above (except for at most 14 days after a teacher link — see “Teacher links” below); whether the app was already set up when this statistics identifier was created; per day (the calendar date on your device), how many times features were used (for example quizzes finished, paywall opened), whether Pro is active, whether an organisation licence is active, the default language pair, library size as a range, number of widgets, whether reminders are on; app version and build number, platform, release channel (App Store, TestFlight or development build) and interface language. No words, no text you typed and no name. The app sends only the calendar date, never a time of day; our server records the time (UTC, by its own clock) it last received each day's counts. That time is not used in our reports and is deleted together with the statistics | Our Cloudflare Worker | To understand which features help people learn and where they get stuck |
AI results are cached on our server, keyed by a hash of the request, so that the same word requested again costs nothing to generate. The cache stores the word and the generated content — not your name, email, or any account identifier. We also keep a count of how often a word is asked for, with no device or account identifier attached to it.
Spoken audio is cached the same way: our server keeps the rendered sound under a hash of the text and the voice, together with the language, the voice name and how many characters it cost — not the text itself, and nothing that identifies you. Your iPhone keeps its own copy in its cache folder, which iOS clears when it needs space, and which “Delete My Data” in Settings also removes.
If you dictate a word, transcription uses Apple's Speech framework. Where your device supports on-device recognition it happens locally; otherwise Apple processes the audio on its servers under Apple's privacy policy. We never receive the audio.
The camera is used to scan Volex QR codes when someone shares words with you, and to photograph a vocabulary page or a book page with the words you marked. A QR scan, and a page photo you use only to type the words in yourself, stay on your device. Only if you choose to have a page read by AI is that photo sent, as described above.
If you enable iCloud Sync, your vocabulary and study progress are mirrored through Apple CloudKit into your own private iCloud database. It is governed by Apple's privacy policy and your iCloud settings. We have no access to it and cannot read it.
Volex Pro is sold through Apple's In-App Purchase. Apple handles the payment; we never see your payment details. We only receive from Apple whether an active subscription exists on your device. If your device is linked to a teacher (see below) and you subscribe, the app also sends the App Store’s signed record of that purchase — the product, its dates, whether it is a free trial and its transaction identifiers, never payment details — so it is counted for that teacher. We keep the subscription’s identifier, product, dates and store environment and the teacher it counts for, not your device identifier.
When you share words as a link or QR code, the content travels in the part of
the URL after #, which browsers never send to a server. Our share
page cannot see what you shared.
A teacher or learning centre may give you a Volex link or QR code. The
teacher’s code travels after the #, so our web page never receives
it. When the link opens Volex, the app sends the teacher’s code, your device
identifier, the platform and the app version to our server; the server keeps
only a one-way hash of the identifier, never the identifier itself. Your device
is linked to that teacher.
The first teacher stays: a device is not moved to another one. Activating an
organisation access code from a teacher’s group links the device the same way.
After one year the hash is replaced by a random value — the totals stay, your
device can no longer be recognised.
Each activation of an organisation access code is recorded with the code, the organisation, the code’s length and the same hash. This is how a trial code is given to a device only once a year.
If usage statistics are on, for at most 14 days after the link the statistics identifier is linked to it, only to count whether the app was actually tried (for example words added or quizzes finished). The link is removed as soon as that is known and after at most 22 days, and at once when you turn usage statistics off or use “Delete My Data”.
A teacher sees only totals — how many devices were linked, tried the app, activated a code or subscribed — never a name, a device or your words. None of this is used for advertising or tracking.
Volex is not directed at children under 13 and we do not knowingly collect personal information from them.
You can request deletion of any data associated with your device identifier by emailing us. Because we hold no account, please send the request from within the app's support flow or include your device identifier. Usage statistics you can delete yourself, at any time, from Settings → Privacy.
If this policy changes materially we will update this page and the "last updated" date above.