What CEFRHub assesses
One production: a written text, or a spoken recording, monologue or exchange.
The same grid applies to French and English: only the markers looked for change, each specific to its language.
Educational assessment tool, not a certifying test. Results are not accepted for any immigration, citizenship or admission process.
The scale: the CEFR and its sub-levels
The Common European Framework of Reference for Languages (CEFR), published by the Council of Europe in 2001 and extended by its Companion Volume in 2020, describes six levels, from A1 to C2, through “can do” descriptors common to all languages.
- A1Breakthrough
- A2Waystage
- B1Threshold
- B2Vantage
- C1Effective Operational Proficiency
- C2Mastery
CEFRHub splits A2, B1 and B2 into two sub-levels: .1, the level reached; .2, the level consolidated. A1, C1 and C2 stay whole. The CEFR itself describes “plus” levels between some levels (A2+, B1+, B2+).
C2 is not native-speaker level: the CEFR says so itself. It describes the precision, appropriateness and ease of highly successful learners.
What is scored
The CEFR describes communicative competence in three components: linguistic, sociolinguistic and pragmatic. CEFRHub's report analyzes all three. The score rests on three marks:
| What is scored | What is observed | Writing | Speaking |
|---|---|---|---|
| Pragmatic competence | Organization of discourse, coherence, cohesion, effectiveness; in speech, managing the exchange | Writing 52 % | Speaking 48 % |
| Linguistic competence | Range and accuracy of vocabulary and structures | Writing 36 % | Speaking 32 % |
| Orthographic (writing) or phonological (speaking) control | Spelling and punctuation; in speech, overall control, sound articulation and prosody | Writing 12 % | Speaking 20 % |
| Sociolinguistic competence | Register, politeness, fit to the situation | analyzed in the report, outside the score | |
In the CEFR, orthographic and phonological control are part of linguistic competence. CEFRHub scores them separately because their weight differs: pronunciation weighs more on intelligibility in speech than spelling does on readability in writing.
The calculation
Each dimension is scored from 1 to 10. The score is their weighted average:
Writing
pragmatic × 0.52 + linguistic × 0.36 + spelling × 0.12
Speaking
pragmatic × 0.48 + linguistic × 0.32 + phonology × 0.20
Transcript of speech, submitted as text
pragmatic × 0.60 + linguistic × 0.40; spelling is not scored
The CEFR prescribes no numerical weighting. This one is an operational reading of the Companion Volume's hierarchy, which positions pragmatic competence as superordinate.
Example: a written text
7.5 × 0.52 + 6.8 × 0.36 + 5.2 × 0.12 = 6.97
B2.1band 6.8 – 7.4
From score to band
| Weighted score | Level | Width |
|---|---|---|
| 2.5 – 3.4 | A1 | 0.9 |
| 3.5 – 4.2 | A2.1 | 0.7 |
| 4.3 – 4.9 | A2.2 | 0.6 |
| 5.0 – 5.9 | B1.1 | 0.9 |
| 6.0 – 6.7 | B1.2 | 0.7 |
| 6.8 – 7.4 | B2.1 | 0.6 |
| 7.5 – 8.2 | B2.2 | 0.7 |
| 8.3 – 9.2 | C1 | 0.9 |
| 9.3 – 10.0 | C2 | 0.7 |
Below 2.5, the production does not yet allow A1 to be awarded; the report says so.
Why the bands are not of equal width
They are narrow at the extremes and wider in the middle, because real progression is not linear either: moving from A1 to A2 asks less than moving from B2 to C1. A wide B1 plateau reflects how long learners actually spend there; a widened B2.2 band reflects how often apparent fluency is mistaken for C1 command.
Safeguards
The score sets a band, not the level awarded. The assessment instructions impose three rules that can lower the computed result:
- Without at least two C1 markers attested in the production, a score of 8.3 or above is capped at B2.2.
- Strong spelling or pronunciation never on its own justifies a move to C1 — the CEFR's “range over accuracy” principle: accuracy without syntactic complexity stays B2.2.
- Each level boundary carries its own exclusion checklist, to be verified before the level is awarded.
The calculation is compensatory; the safeguards are not. A weak competence cannot be offset indefinitely by a strong one.
What the code measures
TTR and MTLD are computed by code, not by the model, on every written text. They are shown in the detailed report, with the word count, for information: computed after the analysis, they do not enter the score.
The role of AI and its limits
The AI model
Reads the production and applies the grid published here: the scores, the weighting, the bands and the safeguards.
The code
Computes the lexical measures and rejects a production that is too short; it does not recompute the level.
The result is an educational diagnosis. A short production, or one far from the task, gives fewer clues: the result is less certain. Data stored and analyzed in Canada.
References
- Council of Europe (2001). Common European Framework of Reference for Languages: Learning, Teaching, Assessment. Council of Europe website
- Council of Europe (2020). CEFR — Companion Volume. Council of Europe website
- McCarthy, P. M. & Jarvis, S. (2010). MTLD, vocd-D, and HD-D: A validation study of sophisticated approaches to lexical diversity assessment. Behavior Research Methods, 42(2), 381–392.
- Halliday, M. A. K. & Hasan, R. (1976). Cohesion in English. London: Longman.
- Brown, P. & Levinson, S. C. (1987). Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press.
- Byram, M. (1997). Teaching and Assessing Intercultural Communicative Competence. Clevedon: Multilingual Matters.
The last three are cited in the detailed-analysis instructions.
Frequently asked questions
Is a CEFRHub level an official level?
No. CEFRHub is an educational assessment tool, not a certifying test: it places a production on the CEFR scale. Results are not accepted for any immigration, citizenship or admission process.
Who assigns the level: an AI or a person?
An AI model, applying the grid on this page. The method was designed by Réda Haiat; the AI Transparency page describes the model's limits and human oversight.
Why is my level lower than my score suggests?
Because the score sets a band and the safeguards can lower it. For example, a score of 8.3 or more stays B2.2 if the production does not show at least two C1 markers.
What do A2.1 or B1.2 mean?
They are CEFRHub's sub-levels: .1 for the level reached, .2 for the level consolidated. A2, B1 and B2 have two; A1, C1 and C2 have none.
Which languages does CEFRHub assess?
French and English, in writing and speaking.