Privacy Policy

Cookie-free traffic counts

We count page loads and views of public learning sections for all visitors, including guests. This counter stores only the UTC date, section name, event type and an aggregate count for up to 90 days. It does not store cookies, visitor identifiers, IP addresses, account details, location or referrers. Counts cannot identify unique visitors. Standard hosting and security processing is separate from this counter.

Google Analytics measurement

On public pages, Google Analytics loads automatically with analytics and advertising storage denied by default. In this limited mode, it sends measurements without analytics cookies to Google. Cookie-free measurement still involves data transmission to Google and is not anonymous simply because cookies are disabled. Google may process request information such as IP addresses to deliver and protect its service. Advertising storage and personalization remain disabled.

You may choose Allow analytics cookies in Analytics settings for fuller visitor measurement. The choice is saved on this browser. We do not intentionally include names, emails, phone numbers, test answers or recordings in analytics events. Account, admin and sign-in pages do not load the tag. Choosing Keep cookie-free stops future analytics cookie storage; it does not stop limited measurements or delete previously collected data. Our separate aggregate traffic counter continues to work independently.

1. Accounts and saved learning data

Sign-in uses Firebase Authentication. Depending on the chosen method, the site receives a verified account identifier, email address or phone number, and an available name. We do not receive your Google password or store your one-time sign-in codes. Historical records created with the former ChatGPT sign-in method remain stored separately; ChatGPT sign-in is no longer offered and those records are not automatically merged into a Firebase profile.

Optional learning preferences include a self-selected reading level, learning goal, up to six interests and a daily study target. These are visible to the organiser. When you enable personalization, the site uses these preferences, saved reading results and structured difficulty feedback to rank reading suggestions. The recommendations use transparent rules and do not require an AI model. When you use the Gemini tutor, enabled personalization also supplies the selected preferences to Gemini; the level and language chosen for the current request take priority.

Signed-in reading submissions also record an attempt identifier, passage identifier, score, question count, per-question correctness flags, a snapshot of saved learning preferences at attempt time, and date. No private written answer text is stored in these reading statistics. Structured feedback such as “Too easy” or “Not interested” is saved per passage. Up to 200 reading-attempt events per UTC day are recorded; latest saved reading scores continue to update.

When you use profile-saving features, our site database stores your account identifier together with practice scores and dates, lesson checks and bookmarks, reading-comprehension scores and dates, shown and answered swipe-question IDs, writing drafts, and saved exam answers or results. This allows you to resume learning and review your progress. The beta organiser can administer this database; saved work is not a public learner feed.

2. Browser storage

The site uses browser local storage for guest swipe history, daily decks, game scores and seen prompts, Sentence Lab history, an active timed test and its answers, mock-question and guided-prompt history, reading-comprehension completion history, and motion preferences. Some game-history keys include the signed-in email address. These records normally remain in that browser until cleared; signing out does not remove them.

To remove device-saved data, clear this website’s storage in your browser settings. Doing so resets device history and can erase unfinished work. It does not remove information already saved to your profile database. Avoid saving private work on a shared device.

3. Recording, voice typing, and playback

Recording requires microphone permission. Ordinary practice recordings are held temporarily in the browser and are not uploaded by the recorder. Download a recording before leaving the activity if you want to keep it.

Tutor voice typing uses your browser’s speech-recognition feature. Depending on the browser, microphone audio may be processed by its speech-service provider. It is separate from the local recorder. You can type instead. Read-aloud uses your device or browser speech service, whose processing depends on that provider.

4. Optional AI processing

The interface indicates whether AI is connected. When tutor chat is enabled and you send a message, the recent conversation, selected level and language, page context, and relevant lesson excerpts are sent to Google Gemini to generate a reply. The current tutor does not save a long-term chat transcript in our database, but records account-linked request entries to enforce usage limits. Chat text remains in the open interface until it is reset or the page is left.

When AI speaking feedback is enabled, a separate consent checkbox is required before uploading your recording and task prompt to Google Gemini for audio analysis, transcription, and feedback. The transcript and assessment are saved to your profile. Our application does not retain the uploaded audio file after processing; provider handling is governed by its own policies.

Our Gemini requests disable API interaction storage with store=false. This does not override Google’s separate logging, retention, or data-use policies. On unpaid Gemini API services, Google may use submitted content and outputs to improve products, and human reviewers may process them. Paid-service terms differ; the organiser must confirm the project’s billing arrangement before inviting users to submit personal information. AI and hosting providers may process data outside your country. Do not submit health, financial, immigration-document, or other sensitive personal information. The site does not automatically turn chats into new teaching content or claim that it trains itself. See Google’s Gemini API terms for its processing practices.

5. Hosting, logs, and analytics

The site runs on OpenAI Sites using Cloudflare infrastructure. Authentication, security, and operational logs may include technical request details handled by these providers. Their processing is subject to their own policies. Our application does not include advertising trackers. After a verified signed-in visit, we keep an account registry with the account identifier, name, available verified email and phone number, sign-in provider, and first/latest verified visit. The authorised organiser can view this registry and saved-progress counts in a private admin dashboard. This registry begins with this release; it is not an import of all Firebase accounts. If additional analytics, advertising, or service providers are added, we will update this notice and seek consent where required.

Optional section usage sharing is off until you choose “Share usage”. If enabled, we record account-linked section names, timestamps and estimated active time while the page is visible and there has been recent interaction. We do not include typed answers, message content, audio, IP addresses, precise location or device fingerprints in these activity records. Charts show participating signed-in accounts only; guests are not tracked by this feature. Recent activity within five minutes is an approximate signal, not proof that a login session is still active. Change this browser’s preference using the usage-sharing controls near the footer. Turning sharing off stops future section tracking; it does not delete past records. Account-registry and saved-learning features continue to work.

The authorised organiser can download a private Excel report containing known account names, contact details, optional country, learning preferences, saved-progress counts and aggregate section activity. The report does not include recordings, private writing, tutor messages or free-text feedback. Download events are recorded with the admin account, export type, row count and date.

A separate AI-improvement permission is off by default and independent of personalization and usage analytics. If enabled, the organiser may export preferences, server-marked reading statistics recorded after that permission began and latest structured passage feedback using new random learner codes for each export. The dataset excludes names, emails, phone numbers, country, account IDs, exact attempt timestamps, recordings, private writing, tutor chats and free-text feedback. Such data is pseudonymous, not guaranteed anonymous. This feature does not automatically train a model or send the dataset to an AI provider. Turn the permission off in My Progress to stop future inclusion. Previously downloaded files cannot be recalled automatically; contact the organiser about access or deletion requests. Re-enabling permission starts a new eligibility period.

You may optionally choose a country or territory in My Progress. The organiser sees country-level counts on a private map and can see the country in your profile row. Map markers show country reference points, not your coordinates. This feature does not collect GPS, IP-derived location, city or street address. Choose “Not shared” to remove the current country selection.

Phone sign-in sends your phone number to Google, which may store it for spam and abuse prevention across Google services. SMS sign-in uses reCAPTCHA and may involve carrier charges. Email sign-in sends a one-time link through Firebase. Neither flow stores the code or sign-in link in the admin dashboard.

Feedback submissions store your account identifier, category, optional rating, message, follow-up permission, date and review status. Your name and email are available to the authorised organiser alongside the message. Feedback is private and is not published to other learners. Checking the follow-up box allows contact about that submission; otherwise it does not authorise follow-up messages.

Google, SMS-code and email-link sign-in use Firebase Authentication. Google/Firebase handles sign-in credentials, account information, and authentication logs. French Gram receives a verified account identifier, available email or phone number, and name, and uses a secure authentication cookie. Firebase maintains sign-in state in your browser. Learning results remain in our site database; they are not sent to a Firebase database. Different Firebase sign-in methods may create separate profiles and are not automatically linked by matching contact details.

6. Retention and your choices

Profile records remain available during the beta until removed by the organiser or a deletion request is handled. There is no automatic expiry schedule or self-service profile-deletion button in this version. Provider backups and operational logs may follow separate retention schedules. We do not promise immediate deletion from all provider systems.

You may request access to, correction of, or deletion of your site-saved data through the contact instructions below. The organiser may ask for enough information to verify account ownership. You may decline microphone permission, avoid AI submission, use guest practice, or stop using the site. Signing out of French Gram does not delete your Google, Firebase or ChatGPT account.

7. Adults, security, and policy changes

This beta is intended for adults aged 18 and over. Do not submit a child’s personal information. Account-specific saving uses authenticated requests; no online service can guarantee complete security. Report a suspected exposure to the organiser.

We will update the date on this page when the policy changes and provide a site notice for material changes in how personal information is used.

Contact during the beta

For support, content corrections, or privacy requests, contact the French Gram beta organiser through the email or message thread in which you received your invitation. If the link was forwarded, ask the sender to connect you with the organiser. Do not send passwords or sensitive documents.

A public support contact will be added before the wider launch.