How the English Assessment works

We assess English skills separately — Reading, Listening and Writing — rather than a single blended score, because a child is often stronger in one skill than another. Speaking is planned for a later phase.

Each skill is reported as an estimated CEFR-aligned level (Pre-A1, A1, A2, and higher as the bank grows), with a confidence indicator and plain-language "I can…" evidence.

Every question is original, written for children and aligned to CEFR concepts. We do not copy content from any commercial test. Items are designed to be culturally neutral and internationally fair, using diverse names and universal topics (family, school, animals, weather, food, travel).

The assessment adapts to the child: consistent success gently raises the difficulty, and difficulty lowers when questions become too hard. One wrong answer never causes a big change — we gather several pieces of evidence before estimating a level, and stop early once the evidence is clear (typically 8–14 short questions per skill).

Writing is assessed by AI (Claude) against level-specific criteria — task completion, vocabulary, language control, range and coherence. The AI assesses only the evidence in the child's writing and never invents evidence. If there isn't enough to judge, we mark it Not assessed rather than guess, and a grown-up can review.

We focus on communication and understanding, not on conforming to one accent or dialect of English. Listening uses clear, intelligible English. When Speaking is added, regional accents will not be penalised.

The assessment is the front door to learning, not the end. Results create a personalised Language Quest, progress is tracked in the child's passport, and a reassessment (suggested after ~8 weeks) shows growth over time.

The Council of Europe provides the CEFR framework; responsibility for the quality and fairness of any assessment lies with those who build it. We are transparent that CEFR alignment is not accreditation. We will publish evidence (such as agreement with teacher assessments) only when it is generated from real data — never as a marketing claim.

Every result stores the assessment version and item-bank version so progress can be compared fairly over time as we improve the content and methodology.