How to get from B1 to B2 (and why you're stuck)

Published: July 9, 2026

B1 is where most self-taught learners stall. You can order food, hold a predictable conversation, follow a slow speaker who is being kind to you. And then progress stops — not because you stopped practicing, but because the kind of practice that got you to B1 can't get you to B2. If you're not sure where you sit on the scale, start with our guide to the CEFR levels; this article assumes you're somewhere around B1 and frustrated.

Why recognition-based drilling stops working at B1

Most language apps are built on recognition: you see four options, you pick the right one, the app celebrates. Recognition is a legitimate skill — it's how you bootstrap a vocabulary from nothing, and it's why those apps work well at A1 and A2. The problem is that recognition and production are different abilities stored differently in your memory. Knowing a word when you see it does not mean you can retrieve it, inflect it, and place it in a sentence under time pressure while someone waits for you to speak.

At B1 the gap between the two becomes the whole problem. Your passive vocabulary is large enough that recognition drills feel easy, so the app keeps serving the same format with slightly bigger words — and none of it transfers to the moment you open your mouth. You keep "learning" words you can't use, and the words you did learn quietly evaporate because nothing forces you to retrieve them. The drills aren't wrong; they're just aimed at a skill you already have. What's untrained is retrieval under pressure, and no multiple-choice format can train it.

What B2 actually requires, according to the CEFR

The CEFR descriptors for B2 are worth reading literally, because they name skills — not vocabulary sizes. A B2 speaker can:

• Produce spontaneously. Interact with enough fluency that a conversation with a native speaker works without strain on either side. No rehearsed phrases; sentences assembled in real time.
• Repair breakdowns. When you're misunderstood — and you will be — rephrase, simplify, give an example, ask a clarifying question, and get the conversation back on track without switching languages.
• Follow native-speed speech. Understand extended speech and follow the thread of a real discussion, including speech that wasn't slowed down for you.
• Defend a viewpoint. Explain a position on a topical issue, give the advantages and disadvantages of options, and hold your ground when someone pushes back.

Notice what's missing: nothing in that list says "recognizes 4,000 words." Every descriptor is about doing something with the language in real time. That's why the jump from B1 to B2 is famously the hardest on the scale — it's not more of the same, it's a different kind of work.

Five practice tactics that actually move you

None of these require an app. All of them require doing the uncomfortable thing on purpose. Where Viglot has an exercise built for the tactic, we'll say so — as one way to execute it, not the only one.

1. Do deliberate speaking reps — with feedback

Speaking practice only counts if something checks whether you were understood. Talking to yourself in the shower builds fluency in the sentences you already own; it doesn't stretch you. What works is a rep with a goal and a verdict: say the thing, find out whether it landed, fix what didn't, repeat. A tutor or a patient friend is the classic setup. A conversation partner who corrects you is better than one who nods along. Viglot's version is Voice Tasks: a spoken conversation with a concrete goal — buy the train ticket, explain the symptoms — where the AI answers in character and then tells you whether you actually accomplished the goal, and what to say better next time. The related skill of surviving being misunderstood gets its own exercise, Make Yourself Understood: the AI deliberately gets you wrong and you have to rephrase and clarify until your meaning lands — the exact repair muscle the B2 descriptors name. More on how this kind of practice works in our guide to AI conversation practice.

2. Listen above your level

If you only listen to material you understand comfortably, you're consolidating B1, not building B2. Native-speed speech is a distinct skill: words run together, intonation carries meaning, and nobody pauses for you. The tactic is to spend part of every week on audio that's genuinely hard — podcasts for natives, films without subtitles in your language — and to test yourself on meaning, not just vibes. Viglot's listening comprehension exercises (dialogue and monologue) do the testing part for you: you follow natural speech, then prove you caught what was actually said. However you do it, the check matters; it's easy to mistake familiarity with the sound of a language for understanding it.

3. Vary your scenarios — around your own life

B1 speakers are fluent inside a small comfort zone: the restaurant, the introduction, the weather. B2 requires handling situations you haven't rehearsed, which means deliberately practicing outside the zone — a complaint, a negotiation, a disagreement about something you actually have opinions on. The "your own life" part matters for retention: vocabulary that connects to your work, your hobbies, your real situations sticks, because your brain has somewhere to put it. Generic dialogues about jobs you'll never work don't. This is the premise Viglot is built on — it generates complete scenarios around what you tell it you care about, at your level — but you can apply the tactic with any tutor by simply bringing your own topics instead of accepting the textbook's.

4. Use spaced repetition for the long tail

The vocabulary you need at B2 is a long tail: thousands of medium-frequency words you'll each encounter rarely. Massed review — cramming a lesson's words the next day and never again — loses exactly these words. Spaced repetition schedules each word's review just before you'd forget it, which is the only time-efficient way to hold a long tail in memory. Any SRS tool works if you feed it. Viglot tracks every word you've seen individually (SM-2 scheduling) and cycles it back automatically, with the reviews demanding production, not just recognition — but a well-kept Anki deck serves the same tactic.

5. Track production, not streaks

A streak measures that you showed up; it says nothing about what you did. You can hold a 400-day streak entirely on recognition drills and stay B1 forever. The metric that predicts progress toward B2 is production volume: sentences you built from scratch, minutes you spent speaking, breakdowns you repaired. Count those. If a week's practice contains no moment where you had to produce language you hadn't prepared, the week didn't move you — however green the calendar looks. This is also a useful filter for tools: ask what an app counts as "learned." Viglot only counts a word once you've produced it yourself — said it aloud, rebuilt the sentence, placed it back in context — which is a design choice you can replicate manually with any method: don't check a word off until you've used it.

The honest timeline

B1 to B2 is typically the longest single jump on the CEFR scale, and no method compresses it into weeks. What a good method changes is the slope: fifteen focused minutes a day of production-heavy practice beats an hour of recognition drills, and most learners who make the switch notice a real difference within a few months. The plateau isn't a fact about you. It's a fact about the kind of practice you've been doing — and it changes when the practice does.