What we collect, and why.

When you sign up with email or Google, we store your email address, display name, and a profile row (level, streak, pet name) used for the gamified progress features. Google OAuth shares only your name and email — we never see your Google password.

Every essay or answer you submit is stored — the prompt, your text, the AI-generated feedback, and the three rubric scores (content, structure, language) — so you can review past attempts and track improvement over time.

We also keep a running list of your recurring mistake patterns (e.g. a grammar point you keep missing), built from your submission history, so drills can target what you personally get wrong.

When you complete a drill, we store what you typed into each blank, the answer the question expected, and whether it was marked right or wrong. Drills are graded by string comparison, not by AI, and nothing here is sent to any AI service.

We keep these for a reason that is about us rather than about you: our question bank lists the answers it accepts, and that list is never complete. An answer several different people gave and were marked wrong for is usually a correct answer we failed to allow for, and seeing those is the only practical way we find them. It also tells us which questions are too hard or badly worded.

If you report a question, we store which question it was, the reason you picked, the answer you had typed, and your note if you wrote one. A report is tied to your account so that one person flagging the same question repeatedly counts once — not so that we can look you up. Reports are read by us to fix the question bank; they are not shown to other learners.

Your submitted text is sent to Anthropic’s Claude API to be graded against the official TOPIK II rubric. See our AI Usage Disclosure for what that involves.

We use three separate layers, each scoped to what it needs:

• Vercel Web Analytics — aggregate visitor and pageview counts, including signed-out traffic. No individual profiling.
• PostHog — anonymized product analytics (clicks, session behavior) used to understand how the app is used and improve it.
• Our own event log — two signed-in-only events (practice started, feedback viewed) tied to your account, used solely to measure completion rates. It holds no text you wrote: not your essays, and not your drill answers or reports, which are stored separately and described above.

We use Supabase (authentication and database) and Anthropic (AI grading) as data processors. We do not sell your data, and we do not share your writing or personal information with anyone else. Row-level security on our database ensures only you (and our service, for grading) can read your own submissions.

Submissions, drill answers and reports, mistake patterns, and profile data are kept for as long as your account exists, so the app can show your history and track your progress. Deleting your account deletes these records.

You can request a copy of your data or account deletion by emailing hello@sseudam.app. Deleting your account removes your profile, submissions, drill answers and reports, mistake patterns, and badges from our database.

If this policy changes in a way that affects how your data is used, we’ll update this page and note the change here.

Last updated: October 2026.