English level test
How the test works
How the test works
The test is adaptive. It starts in the middle, and after each short block of questions it updates its estimate of your level and chooses the next questions to fit it. That's why two people never get the same test, and why the test can be short and still precise.
It has four parts:
- Words. Grids of real and made-up words. Ticking a made-up word tells us you're guessing, so the score allows for it.
- Grammar. Gap-fill questions with three options, typed answers, word order, word formation and sentence rewriting.
- Listening. You type short sentences you hear, and answer questions about short recordings.
- Reading. Short texts with the second half of some words missing, and questions about short texts.
Every answer is checked by fixed rules: a key, a list of accepted answers, or a count of correct words. No AI decides your result.
From answers to a level
Each question has a measured difficulty on one scale. Your level is the point on that scale that best explains your answers, together with how sure we are of it. When your answers fit two levels, the result says "close to" the other level. Your score from 100 to 699 is the same estimate in finer steps: the first digit is your level (1 = A1 … 6 = C2).
The levels follow the CEFR, the Council of Europe's six-level scale. The link between our scale and the levels comes from the questions' own levels, checked against the results of learners with known levels and exam scores. It is our estimate; it is not endorsed by the Council of Europe.
How questions are made and checked
Every question is written for this test. Before anyone sees it, automatic checks make sure each word is at the right level, that independent solvers who don't see the answer all agree on it, and that no wrong option can be argued for. Most questions count from the start at the level they were written for; real answers then measure how hard each one really is, and results are re-scored as those measurements come in. Some new reading and listening questions appear without counting until real answers show they work. Questions that misbehave are removed automatically, and the "Something wrong with this question?" link on every question helps us find problems faster.
During the first weeks, while questions are still being measured, results are marked "provisional" and may be updated.
Limits
- The test doesn't score speaking or writing. You can rate those yourself at the end.
- It's an online test with no supervision, so it can't confirm who took it or how.
- The quick test is less exact than the full test.
- It isn't an official certificate. For a visa, a university or a job, take an official exam.
Retakes and fairness
Each question exists in several versions, and the test avoids showing you a question you've seen before. We never show the answers to scored questions, so there's nothing to memorise, and questions that start to look memorised are retired.
Credits
- Word levels: the CEFR-J Wordlist 1.5 and Grammar Profile (Yukio Tono, Tokyo University of Foreign Studies, cefr-j.org) and the Octanove Vocabulary Profile C1/C2 (CC BY-SA 4.0).
- Word frequencies: wordfreq by Robyn Speer and contributors (data CC BY-SA 4.0).
- Listening voices are synthetic, made with Kokoro-82M (Apache 2.0).
An estimate aligned to the CEFR by Englishery. Not an official certificate, and not endorsed by the Council of Europe.