Privacy Policy

When you sign in, we collect:

• Email address — from Google Sign-In or Apple Sign-In (OAuth provider)
• Display name — from OAuth provider, editable by you
• Profile photo URL — from OAuth provider, editable by you
• Unique user identifier (UID) — generated by Firebase Authentication
• Preferred language — auto-detected from device locale (BCP-47 format)
• Account creation timestamp
• Last active timestamp — updated with 5-minute debounce
• Online status: while the app is open, your device writes a small presence record (online, in a match, or offline, plus a last-seen time, whether that match needs your full attention, and whether your Live play setting is on) about every 2 minutes, or about every 25 seconds while a Live invitation you sent is open. Only players you have accepted as friends can see it, as a green or amber dot next to your name and to know whether they can invite you to a Live duel; it is never shown to other players. Turn off Online status in Settings, Privacy to stop publishing it (you then stop seeing your friends' dots too). The record is deleted with your account.
• Onboarding (before sign-up): when you first open the app, we create a temporary anonymous account (a random Firebase UID) so you can try LearnClash before signing up. It is merged into your account when you sign up, or deleted automatically if you do not.
• Sign-in restore key (Android): when you are signed in on Android, your device stores a restore key with Google Play services so that, when you move to a new Android phone and restore your data, you are signed in to LearnClash automatically. We store only the public half of the key together with your user ID; the secret half never leaves Google's systems. The key record on our servers is deleted when you sign out or delete your account; uninstalling the app removes only the copy on your device.

To provide matchmaking and track your progress:

• Skill rating — matchmaking score (Glicko-2 system, shown as an ELO-style number) starting at 1300
• XP (experience points) — progression tracking
• Duel count — total games played
• Current streak — consecutive days you've been active
• Maximum streak — your all-time best
• Streak freezes — available streak protection items
• Last streak date — ISO format (YYYY-MM-DD)
• Preferred categories — your 3-18 category selections for matchmaking
• My topic IDs — topics you're practicing

To implement spaced repetition learning:

• Per-question correct count — how many times you answered correctly
• Last seen timestamp — when you last saw each question
• Next review date — scheduled by our SRS algorithm
• Topic cooldown timestamps — prevents overuse of same topics

For each game you play:

• Duel ID and timestamps — created, updated, expires
• Opponent information — UID, display name, photo, ELO (public profile data)
• Round-by-round answers — which option you selected and correctness
• Response time — milliseconds to answer each question
• XP earned — per answer
• Final scores and ELO changes
• Forfeit status — if applicable
• Answer touch characteristics: for each answer you tap in a rated duel, the pointer type, pressure, contact size, hold duration and position on the option, together with your device's operating system and model. Used only for fair-play checks and never to identify you or to link accounts.

For friend features and multiplayer:

• Friend relationships — user IDs, pending/accepted status, friend streak
• Head-to-head statistics — win/loss record against specific opponents
• Unread notification count
• Public profile — your display name, photo, ELO, and XP are visible to other authenticated users for matchmaking and leaderboards
• Question reports — if you report a question issue, we store your user ID with the report

When you send feedback or a bug report from the app, we collect only what is needed to understand and route the report:

• Message and report type - the text you enter and whether you selected bug or idea
• Optional app status diagnostics - app version, build number, flavor, platform, OS version, device model, locale, timezone, text settings, layout insets, network state, route label, and progress counters such as streak, ELO, Mems, level, and XP
• Optional screenshot - captured only after you open the feedback sheet, shown as a preview, removable before sending, and disabled on private screens and active gameplay screens
• Support routing metadata - report ID, Firebase Auth UID, privacy hash, support delivery status, attempts, and error code if delivery fails
• Data we do not include - raw logs, notification tokens, full question content, passwords, payment details, or public screenshot URLs

For app functionality and security:

• FCM tokens — Firebase Cloud Messaging tokens for push notifications (max 5 per user)
• Firebase App Check attestation — device integrity verification
• Privacy hash — HMAC-SHA256 of your UID for anonymized crash correlation
• Installation UUID — Crashlytics installation identifier
• IP address — processed transiently by Firebase infrastructure, not stored by us

To improve the app:

• Screen views and navigation paths
• Feature interaction events — duel created, question answered, etc.
• Session duration and timestamps
• Paywall views, conversions, dismissals
• Error logs and stack traces — via Firebase Crashlytics
• Diagnostic keys — FCM status, APNs availability
• Breadcrumb logs — activity trail preceding crashes

Google Analytics Advertising Features and User-Provided Data

When you grant ads, demographics, and matching consent on our website, Google Analytics may additionally collect or receive:

• Demographics — inferred age range and gender
• Interests — inferred interest categories based on browsing history
• Cross-device linkage — associates activity across devices you use while signed in to the same Google account
• User-provided data — after website sign-in, a SHA-256 hash of your normalized account email may be sent to Google Analytics for enhanced conversions, Customer Match, demographics, interest reporting, and attribution. We do not send raw email, names, phone numbers, street addresses, city, postal code, or precise location.

These features are disabled by default in the EU/EEA/UK/Switzerland and only activate after you click "Accept all" or opt in via the Privacy preferences link. Global Privacy Control also disables them. You can opt out at any time via Google's Analytics Opt-Out Browser Add-on, your Google Ads Settings, or by clicking "Privacy preferences" in our footer. In the mobile app: we use Firebase Analytics and share measurement data with Google Ads to measure which ads lead to installs and trials and to improve our advertising. On iOS, advertising measurement is on by default outside the EU/EEA/UK/Switzerland and stays off until you opt in within them. If you also allow analytics and are signed in, we use your sign-in email with Google's on-device conversion measurement. This helps measure and improve our Google Ads without your identifiable email leaving the device through this measurement. This is separate from the website's hashed-email sharing described above. On Android, advertising measurement is on by default outside the EU/EEA/UK/Switzerland and stays off until you opt in within them. On both platforms, you can change either purpose anytime under Settings, then Analytics & Ad Measurement. Analytics has its own separate choice and regional defaults. On Android, the same setting also controls AppsFlyer, an install-attribution service that measures which ad networks lead to installs and trials; AppsFlyer receives the Google Advertising ID, an AppsFlyer-generated device identifier, IP address, device information, install events, and trial/subscription events (no payment details, no email). The app does not use Apple's advertising identifier (IDFA) and, on iOS, does not track you across other companies' apps or websites.

Managed by RevenueCat:

• Purchase history and product IDs
• Subscription status — active, expired, grace period
• Store identifier — App Store, Google Play, or web (Stripe)
• Expiration and renewal dates
• Web purchases: subscriptions bought on learnclash.com use RevenueCat Web Billing with payment processing by Stripe. Your card details are entered directly with Stripe and are never stored by us.

If you use AI chat or duel chat features:

• AI conversation history — automatically deleted after 30 days (TTL)
• Duel chat messages — message text, sender user ID, duel ID, thread ID, timestamps, and moderation status
• Clash AI prompts — when you ask Clash or tag @clash, the text you submit is processed to generate the reply. Duel-thread Clash replies use only your visible @clash message and do not include hidden duel state
• Chat safety settings — terms acceptance, chat preference, safe mode, muted or blocked users, and local hide records
• Report and moderation records — created when you report a message or when automated safety systems block a message

The LearnClash Chrome Extension stores the following data locally on your device:

• Cached rounds and recent question history: up to 24 unused rounds and 256 issued-round records within storage limits. They are usable for seven days. Expired entries are no longer used and are removed when the cache is next updated.
• Preferences and progress: language, difficulty, daily streak, last completion date, completed-round count and rating-prompt preferences.
• Installation identifier: a randomly generated identifier used to group usage events. It is not linked to a LearnClash account.

The extension sends basic usage events through the LearnClash API to Google Analytics 4 (GA4) using the Measurement Protocol. These events are associated with pseudonymous installation and session identifiers:

• Quiz activity: extension opens, assigned topic name and ID, difficulty, answer correctness, question position, completed-round score and interactions with extension links and controls. Search use is counted, but search terms and selected answer text are not sent to LearnClash.
• Reliability and engagement: extension version, cache use, loading duration, failure category and time the extension page is visible.
• Session identifier: a numeric identifier shared across extension pages, renewed after 30 minutes of inactivity or a browser restart.

Quiz and analytics requests also process the following operational information:

• Language, difficulty and question-bank selection: used to return suitable quiz content.
• Request timestamp: checked for request freshness.
• IP address and request metadata: processed for delivery, security and rate limiting. IP-based counters are held temporarily in server memory, and request metadata may appear in infrastructure logs.

No account required: the extension does not ask for your name, email or app account, and does not read your browsing history or the content of visited pages. Searches go through Chrome to your configured search provider. Uninstalling removes the extension's local data; it does not automatically erase usage events already sent to GA4.

If you create a private quiz topic from your own study material, we process only what you add and only to build and moderate that topic:

• Material you add - pasted text, web links, YouTube links, PDF files and photos of notes. Files are uploaded to private Cloud Storage. Our backend extracts embedded PDF text directly; Google Cloud Vision processes every PDF page for text recognition (OCR) and image safety, including scanned pages. Gemini receives the extracted PDF text. If extraction indicates unclear formulas, handwriting or page layout, we also send images of selected PDF pages to the paid Gemini API to clarify the source material. We do not send the original PDF file to Gemini or store separate copies of these page images; the resulting text is kept with the extracted material. For photos, our backend grants the paid Gemini API access to each file for extraction and requests removal of that file registration after each attempt. The API fetches the original photo without keeping a separate uploaded-file copy. We delete our originals after extraction; unfinished originals have a 7-day automatic clean-up. Google's separate abuse-monitoring retention still applies.
• Extracted text - the text taken from each source, stored with the topic until you delete the topic or your account, because every later generation run and every fact check of a reported question needs it.
• Screening results - an automated image-safety check for photos and every PDF page and an automated content screening for every source (safe or not, whether it contains instructions to the AI, whether it contains other people's personal data, a one-sentence subject summary, and the detected language). Rejected material is not kept in LearnClash topic storage; Google's abuse-monitoring retention still applies.
• Topic and question data - the topic name and category you chose, the questions generated from your material, question counts, and per-source content hashes (used only to detect the same upload being added twice)
• Consent record - the version of the content rules and AI-processing notice you accepted and when
• Cover image - an AI-generated cover created from the topic name and category only, never from your material
• Who can see what: your material and its extracted text are never shown to other users. Only you (when you review extracted text in the app) and, for moderation and support, authorized LearnClash staff can access them. The processors listed in Section 4 (Firebase, the paid Gemini API and Cloud Vision) store and process them on our behalf under their applicable data-processing terms. Google's abuse-monitoring retention and review are described in Sections 4 and 5. Friends you challenge see the topic name and generated questions, never the material. Nothing from a study material topic enters the public topic pool or search.