How FrenchExaminer scores your French

We measure first, then explain. A speech-scoring engine measures your pronunciation word by word; an examiner model explains only what was measured and corrects your grammar and vocabulary; your level is always shown as a range on the NCLC scale, never as an official score.

Pronunciation: measured, not guessed

Your recording goes through a speech-assessment engine that scores each word. Words under 80/100 are marked as words to work on. Words the engine did not understand clearly are marked as such, never as mistakes. Research shows that general AI models asked to judge an accent tend to invent errors; that is why the model only explains scores it receives.

Speaking: the official tasks and criteria

The speaking simulation follows the three TCF Canada tasks with their timings. Each answer is assessed on the task, fluency, vocabulary, grammar and pronunciation, then turned into an estimated NCLC range.

Writing: criterion by criterion

Each text is read against the official criteria: task completion, coherence, vocabulary, grammar and spelling, and the required word count. You get the sentences to fix with the reason, and an improved version of your own text.

A second, independent reading

A share of writing corrections is graded twice by independent readings. When they disagree by more than one level, the range we show you is widened instead of hiding the doubt.

Checked against official results

When candidates share their official TCF or TEF results, we compare them with our estimates. Once enough results are in, we will publish the gap on the home page.

What our estimate is not

It is not an official result and it does not replace the exam. One short answer gives a wide range; several full simulations give a narrower one. We say so on every report.