How scoring works

A useful practice score should explain its evidence and its limits. TestDayTwin uses three different scoring paths because not every language skill can be measured with the same confidence.

1. Reading and listening: answer-key scoring

Objective questions are checked against authored answer keys and accepted alternatives. Fill-in answers normalize safe differences such as capitalization and explicitly accepted spellings. Raw totals are converted only where the exam publishes a usable conversion. This is the most dependable part of the system, although an unofficial question can never predict an official result.

2. Writing: local practice estimates

Standard writing feedback runs in your browser. It checks measurable features such as word count, required form, spelling, grammar, vocabulary range, sentence patterns and overlap with the task. The full grammar checker also runs locally and loads only after you start writing.

Strongest signals: length, hard form rules, spelling and grammar patterns. Roughest signals: task response, idea quality, relevance and whether an argument would persuade a human examiner. The interface labels rough criteria instead of presenting them as facts.

3. Optional AI deep-score

If you explicitly choose detailed AI scoring, the task prompt, your submitted response and a small set of numeric writing measurements are sent through TestDayTwin's protected endpoint to Google Gemini. The response is returned to your browser and TestDayTwin does not store the essay or the generated feedback.

The result shows a calibration label. Uncalibrated means rough guidance only; provisional means checked against a small examiner-scored sample; validated is reserved for a stronger held-out evaluation. No label turns a practice estimate into an official score.

What progress sync contains

When signed in, progress sync stores the selected exam, goal, practice-page path, skill, completion status, score summary and timestamp. It does not contain question answers, essays, recordings, question text or AI feedback. See the Privacy Policy for deletion and retention details.

How to use the result

Use objective scores to identify weak question types and timing problems.
Use writing criterion feedback to choose one concrete improvement for the next response.
Look for a trend across several attempts; do not treat one estimate as a verdict.
Use official exam materials and qualified human feedback for high-stakes decisions.

Independent practice, not an official exam

TestDayTwin is not affiliated with or endorsed by the exam owners. Its scores are practice estimates and cannot be used for immigration, admission, licensing or employment.