Effective October 21, 2026

Privacy policy

Parlance is a vocabulary app. This policy describes everything the app does with data. It is short because the app does little with data, and we intend to keep it that way.

No account

There is nothing to sign up for. The app does not ask for your name, email address, or Apple ID. Requests do carry a random installation identifier, so we can count requests from the same installation. Text you submit may itself contain personal information; avoid pasting sensitive or confidential material.

What stays on your device

• A downloaded copy of your word list; our server also caches the list.
• Your progress — which words you've seen and marked as used.
• Your settings, including the accent you chose for pronunciation.
• Your notification schedule. Words are delivered by local notifications, scheduled on the device itself. There are no push notifications and no server involved in delivery.

This data is included in your device backups under Apple's normal backup rules, and it is deleted when you delete the app.

What leaves your device

Source and pronunciation requests go to our server. App verification also contacts Firebase automatically, as described below.

Choosing a source. When you pick a source, the app sends your choice to our server to generate your word list. If your source is a TV show, a book or a podcast, that request contains the title you picked. If your source is your own text, the request contains the text you pasted.

Hearing a word. When you tap the speaker button, the app sends the word, its phonetic spelling, and one of the app's own example sentences to our server, which has it spoken by Google Cloud Text-to-Speech. The clip is cached on our server so the next person to hear that word does not wait. If the server cannot be reached, the speech synthesiser built into iOS speaks the word on your device instead. We retain request counts for rate limits, not a listening history in your word-bank record.

Both requests carry a random identifier the app creates on first launch and stores on your device. Our server retains that identifier with request counters to limit how many word lists and clips one installation can request per day. It is not your Apple ID or an advertising identifier. Submitted content and request counts are linked to the installation when received, even though there is no account.

What our server keeps

• The word list it generated — the words, their definitions, and, for pasted text, the short passages (a sentence or less) that were matched against your text to verify a quotation. Your full text is not stored. The list is kept so that anyone who pastes the identical text, including you on a new device, gets the same words without waiting. The cached bank has no installation-ID field and is keyed by content, not by a person. Banks have no automatic expiry. Deleting the app does not delete the server cache.
• For a TV show, book or podcast, the title. A title that is not already in our catalogue is added to it, so other people can pick it too. The words are the same for everyone who chooses that title.
• Cached pronunciation clips, keyed by word and voice — not by who asked.
• Installation identifiers and request counters. Each daily counter expires two days after its UTC window starts; a daily cleanup removes expired records. With cleanup running normally, a record can remain for up to three days from the start of its window.
• Operational logs and sampled traces for troubleshooting: request metadata, generation timing, errors, and media titles. We do not intentionally log pasted text. Cloudflare's configured log retention applies separately from the rate-limit cleanup.

Service providers

Our server runs on Cloudflare, which processes requests to deliver them to us and to block abusive traffic; Cloudflare sees the network information any web server sees. Words are selected and defined by Google's Vertex AI (Gemini), and spoken by Google Cloud Text-to-Speech. Your input is processed for those purposes only. We do not use it to build a profile of you, we do not use it for advertising, and we do not sell it. Google processes it under the Google Cloud privacy notice, and Vertex AI does not use it to train models. Firebase App Check uses Apple's App Attest to verify that requests come from a genuine copy of the app. Firebase receives attestation and assertion objects and returns a token the app sends to our server.

Firebase App Check also sends device model, OS and SDK/build-version information with daily heartbeat dates. Google uses this information to measure platform and version adoption and inform Firebase development. Google states that this Firebase user-agent information is not linked to a user or device identifier. See Firebase's data-collection disclosures.

What we don't do

• No advertising tracking or behavioral analytics SDK. Firebase's technical adoption measurement is described above.
• No advertising, and no advertising identifiers.
• No selling or sharing of data with third parties, other than the service providers named above.
• No cookies on this website, and no third-party requests from it — even the fonts are served from our own domain.

Children

Parlance is not directed at children under 13. If you believe a child has submitted personal information, contact us at the address below.

Changes

If this policy changes, we will update this page and the effective date above. A material change — anything that sends more data off your device than described here — will also be called out in the app's release notes.

Contact

Questions about this policy: support@useparlance.com.