English level test: how to check your real level with the CEFR "I can" grid

A free English level test gives you an estimate of what you recognise when you read and listen, but it never hears you speak. To find out your real speaking level, use the official CEFR self-assessment grid, which describes each level from A1 to C2 as "I can" statements, then test yourself out loud by recording a real task and comparing it with the descriptors (Council of Europe). The result tells you where you stand and what to practise. It is not a certificate.

You sit an English level test online one evening, fill the gaps, answer two audio clips, and the screen says B2. A week later, in a job interview in English, the question "tell me about a time you disagreed with your manager" leaves you hunting for words. The test measured one thing and the interview asked for another. This guide gives you the official, free yardstick for the part the test cannot hear.

English level test: what an online test measures and what it misses

Almost every free English level test works the same way: questions on screen, options, gaps, a short recording. That format measures recognition: you see the right answer among three and pick it.

Serious placement tests work like this too. The Cambridge English Placement Test is a computer-based online test of reading and listening, with questions answered on screen, including multiple-choice and gap-fill items (Cambridge English). According to Cambridge's help centre, it takes about 30 minutes, is marked automatically, gives results instantly and was designed to test levels A1 to B2 most robustly, because that is where most candidates fall (Cambridge English). Even a well-built test leaves speaking out by design. We mention it as an example of the format, not as a recommendation.

The CEFR itself treats understanding and speaking as separate skills. On the official grid, B2 listening means you can follow lectures and complex lines of argument on reasonably familiar topics and understand most films in standard dialect (Europass). That row says nothing about answering back. Speaking has rows of its own.

And it is normal for the rows not to match. The CEFR Companion volume, published in 2020, describes an adult who is far stronger in reading and says that almost every learner's proficiency will be uneven and partial (Council of Europe, 2020). Your B2 from the online test can be true for listening while your conversation sits at B1. The difference lies in what each one measures.

The official CEFR "I can" grid

Search for a free English level test, A1 to C2, and you rarely find Table 2 of the CEFR, the self-assessment grid. The Council of Europe calls it an orientation tool to help learners profile their main language skills and decide at which level to look at more detailed descriptors (Council of Europe). It separates listening, reading, spoken interaction, spoken production and writing, and Europass, run by the European Commission, publishes it free, without registration (Europass).

Two rows matter for speaking: spoken interaction (talking with someone) and spoken production (speaking on your own for a while, describing, telling, arguing). Here they are, taken from the official English text and shortened where marked [...]:

A note on B2. The 2001 wording speaks of regular interaction "with native speakers". The 2020 edition proposed changes to descriptors that refer to "native speakers", because the term has become controversial since the CEFR was first published (Council of Europe, 2020). For the same reason, C2 is not "native level": the framework states that C2 has no relation whatsoever with the performance of an idealised native speaker.

How to use the grid

Read the interaction column from the top and stop at the last row where you can honestly say "yes, without help". Then do the same, separately, for production: you may land on a different level in each. The grid has no score and no official adding-up rule. If you are torn between two rows, our suggestion is to take the lower one and treat the higher one as your target.

If you have to state a level on an exchange or university application form, the grid helps you choose honestly. Check first whether the institution accepts a self-declared level or asks for a certificate.

The 20-minute test: record yourself and listen like an examiner

Ticking the grid from memory has a flaw: you remember your best conversation, not your average one. So here, memory gives way to a recording. Twenty minutes is the length we suggest for the whole exercise; it is not a figure from any source.

Pick your scene: meeting, phone call or interview

The CEFR Companion volume has dedicated scales for these three situations (Council of Europe, 2020):

• Meetings: at A2 you say what you think when addressed directly; at B1 you put over a point of view clearly but find it hard to engage in debate; at B2 you keep up with an animated discussion; at C1 you disagree diplomatically.
• Phone and video calls: at A2, a short, simple call with someone you know on a predictable topic; at B1, everyday personal or work use, asking for clarification from time to time; at B2, a range of work purposes, provided you can ask for clarification when an accent or a term is unfamiliar; at C1, most professional purposes.
• Interviews, including job interviews: at B1 you take some initiative but depend heavily on the interviewer; at B2 you take the initiative and develop ideas with little prompting; at C1 you take part fully, without support.

Speak for 2 minutes without a script and record it

Open the voice recorder on your phone. Take a task from the spoken production row of the level you ticked: at B1, describe an experience and briefly give reasons for an opinion or a plan; at B2, explain a viewpoint on a topical issue with the advantages and disadvantages of each option (Europass).

A B2 task for the interview scene: "Tell me about a decision at work or university you disagreed with, and what you would have done instead." Speak without notes and without pausing the recording. The 2 minutes are also our suggestion: long enough for the rehearsed sentences to run out and your unprepared English to show.

Listen with the Table 3 criteria

Table 3 of the CEFR was designed to assess spoken performance and looks at five aspects: range, accuracy, fluency, interaction and coherence (Council of Europe). Listen to the whole recording, then again with these questions, drawn from the official descriptions in that table:

• Are your utterances very short, with pauses, false starts and reformulation that are very evident? That is A2 fluency.
• Do you keep going and stay understandable, while pausing to plan grammar and vocabulary and to repair mistakes is very evident in longer stretches? That describes B1.
• Is your tempo fairly even, with some hesitation as you search for patterns but few noticeably long pauses, and do your errors never cause misunderstanding? That is B2.
• Do you speak fluently and spontaneously, almost effortlessly, with only a conceptually difficult subject breaking the flow? That is C1.

Your result checklist

Go back to the grid and, for each speaking row, choose one of three answers:

1. I do this without help;
2. I do this with help or a lot of hesitation;
3. not yet.

A hypothetical example. Tom scored B2 in listening on an online test. In his recorded interview answer, he described the situation clearly but stalled when explaining what he would have done instead. On the checklist, B1 interaction was "without help" and B2 was "with hesitation": B2 listening, B1 conversation. That does not make Tom "a B1". It means his comprehension is ahead of his speaking, and conversation is what he needs to train next.

How far can you trust your self-assessment?

A meta-analysis in the journal Language Testing pooled 67 studies with more than 68,500 participants and found an overall correlation of .466 between self-assessment and measured language performance (Li and Zhang, 2020). That is a moderate relationship: self-assessment points in the right direction but does not hit the exact spot.

The study found factors that change the result, including the type and presence of criteria, the instrument and training. The abstract does not say how much each one helps, so we promise no precision. That is why the method above rests on written criteria, the official grid, and recorded evidence instead of memory.

Self-assessment helps you find your bearings and choose what to practise. It proves nothing to anyone else.

If you need to prove your level for a job, a university or a visa, you need a certified exam, with its own scoring rules. How to show your level on a CV is a topic for another guide.

Once you know your level: train the row that lags behind

Speaking improves by speaking. More multiple-choice tests mostly train the row that was already ahead. Take the grid row just above yours (our guide to what the CEFR levels really mean when you speak shows how each one sounds) and make it your target: the grid was built precisely to help you decide which level to look at next (Council of Europe).

After a few weeks of practising out loud, repeat the recorded test with the same scene and compare the two recordings. If the pauses are shorter and you made your point without rehearsing it, the row has moved. How long B1 to B2 takes is a topic for another article in this series.

To practise conversation, you need someone on the other side. The Alyx AI Teacher for language learning talks with you and corrects pronunciation, grammar and vocabulary in real time, during the conversation, without interrupting it, in six languages and from A1 to C2.

A scene to rehearse out loud

A job interview in English, set up as a conversation with the AI Teacher. The target is the B2 interview row: take the initiative and develop your ideas with little prompting.

• AI Teacher: "Tell me about a time you disagreed with a decision at work."
• You: "Last year my team wanted to cut user testing to save a week. I thought that was risky, so I asked for two days instead of five."
• AI Teacher: "And how did your manager react?"
• You: "She was sceptical at first, but I showed her two bugs we'd missed in the previous release, and she agreed to try it."

Record your answers, listen with the Table 3 questions, and try again with a different story. When you can develop the answer without being prompted, you are doing what the B2 descriptor describes.