CEFR speaking test: how CEFRHub assesses your recording

How does CEFRHub assess a recording?

In short: You record yourself on the spot or upload an audio file, monologue or exchange. The recording receives three marks out of 10, combined into a weighted average that falls into a score band; safeguards check the level boundary before the level and the report are awarded.

1. 1Choose speaking and the assessed language, French or English, then a target level to get a prompt at that level, or free production to speak on a topic of your choice.
2. 2Record yourself on the spot, with the browser's microphone, or upload an audio file in MP3 or WAV format.
3. 3Read the analysis, which arrives in under 3 minutes: the level and sub-level, the descriptors that apply to the recording, the strengths and what is left to work on.

One production: a written text, or a spoken recording, monologue or exchange.

Behind it, the grid is published: three marks out of 10, a weighted average, a score band, safeguards at each level boundary, then the level and the report. An AI model reads the production and applies the grid published in the methodology: 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.

CEFRHub is an educational assessment aligned with the CEFR: it is neither an official exam nor a certification, and it is not a conversation partner. It assesses spoken production; it does not measure listening comprehension.

What is scored, and how much does each part weigh?

In short: Three marks: pragmatic competence (48% of the score), linguistic competence (32%) and phonological control (20%). Sociolinguistic competence (register, politeness, fit to the situation) is analyzed in the report, outside the score.

Each dimension is scored from 1 to 10. The score is their weighted average: pragmatic × 0.48 + linguistic × 0.32 + phonology × 0.20.

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 CEFR prescribes no numerical weighting. This one is an operational reading of the Companion Volume's hierarchy, which positions pragmatic competence as superordinate.

A transcript of speech submitted as text is scored differently: pragmatic × 0.60 + linguistic × 0.40; spelling is not scored.

How does the score become a CEFR level?

In short: The weighted average falls into one of the nine published bands, from A1 to C2. The score sets a band, not the level awarded: three safeguards can lower the computed result.

Below 2.5, the production does not yet allow A1 to be awarded; the report says so.

Safeguards

• 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.

In the CEFRHub assessment, the sub-level places your production within the level: .1, the level is reached; .2, it is consolidated and features of the next level are starting to appear, what the CEFR calls a “plus” level (A2+, B1+, B2+). A2, B1 and B2 have two sub-levels; A1, C1 and C2 have none.

How long should you speak?

In short: From 30 to 45 seconds at A1 to 2 to 3 minutes at C2. A recording under 15 seconds is not assessed; with the free account, you can record up to 3 minutes.

A recording under 15 seconds is not assessed.

With the free account, you can record up to 3 minutes: the longest expected duration in the bank, 2 to 3 minutes, fits.

• The result is an educational diagnosis. A short recording, or one far from the task, gives fewer clues: the result is less certain.
• Near the boundary between two levels (e.g. B1/B2), the AI may under- or over-estimate the level.
• Record yourself somewhere quiet: background noise gets in the way of the analysis.
• A text read aloud or learned by heart does not show the same thing as spontaneous speech: the tool asks you for the production type to take this into account.

What does the report contain?

In short: The level and sub-level on the CEFR scale, a summary, an analysis of three competences with examples taken from the recording, pronunciation included, strengths, points to improve, recommendations and the next level.

• The level detected and the sub-level (for example B2.2) on the CEFR scale.
• A summary of the production and of what places it at this level.
• The detailed analysis of three competences: linguistic, sociolinguistic, pragmatic, with examples taken from the production.
• Strengths, points to improve and recommendations.
• The next level and what it requires.

For a recording, linguistic competence includes the phonetic and phonological analysis (pronunciation, intelligibility).

The free account gets a basic report (about 500 words) and a 2-page PDF summary; the full report in the sample corresponds to the paid plans.

What does the CEFR say about spoken production?

In short: Six scales of the Companion volume cover oral production. CEFRHub's report draws on the descriptors of the CEFR and its Companion Volume, which it quotes.

Here, quoted from the Companion Volume, is what the overall oral production scale describes at B1, B2 and C1.

• B1

  Can reasonably fluently sustain a straightforward description of one of a variety of subjects within their field of interest, presenting it as a linear sequence of points. (p. 62)

• B2

  Can give clear, detailed descriptions and presentations on a wide range of subjects related to their field of interest, expanding and supporting ideas with subsidiary points and relevant examples. (p. 62)

• C1

  Can give clear, detailed descriptions and presentations on complex subjects, integrating sub-themes, developing particular points and rounding off with an appropriate conclusion. (p. 62)

Quoted descriptors: © Council of Europe, 2020. Council of Europe — CEFR Companion volume, English edition, 2020 (PDF)

What is the teacher's role?

In short: CEFRHub is for a first reading and low-stakes decisions: placing a learner in a group, tracking progress. It does not take the final decision: the teacher reviews what counts.

• A first reading in minutes, to free time for feedback and teaching.
• Low-stakes decisions: placing a learner in a group, tracking progress between two assessments.
• The same published method for every production, which makes it easier to compare a learner with themselves over time.

• CEFRHub is not a certifying test: when an official result is required, the learner needs an official exam.
• CEFRHub does not take the final decision: the teacher reviews what counts.
• CEFRHub does not assess comprehension: it assesses written and spoken production.

Is it free to assess a recording with CEFRHub?

In short: The first spoken assessment is free, once, with no credit card. After that, each spoken assessment uses one credit, from CA$5, or is included in a subscription.

Free to start, no credit card: one spoken assessment offered once and one written assessment every 30 days. After that, credits with no subscription from CA$5 (1 credit = 1 assessment), or a subscription: Professional at CA$49/month (50 assessments), Power at CA$89/month (200 assessments), Expert at CA$179/month (unlimited assessments, under a fair-use policy: 50 analyses per day). 7-day money-back guarantee on the first subscription (full refund if 5 assessments or fewer); cancel anytime, with no cancellation fees. Plus applicable taxes.