LingQ built one of its most-loved features around a single running number: known words. Learners screenshot it, compare it, watch it climb. It works because it turns something usually vague — "I'm getting better at Spanish" — into something concrete you can point at. LingoBlend now has its own version of that number, and we built it to answer the honest question rather than the flattering one.
What actually gets counted#
Not every saved word counts, and that's deliberate. A word you added to your dictionary yesterday and haven't reviewed yet tells you nothing about what you actually know — it tells you what you intend to know. So the counter only credits words that have graduated: ones that have passed the initial learning phase in spaced repetition, meaning you've answered them correctly enough times, across enough sessions, that the app has stopped treating them as brand new. A word can graduate and still be shaky — "mastered" is a higher bar again — but graduation is the line between "on your list" and "you've actually met this word and it stuck."
That's a meaningfully stricter bar than most apps use. A raw save count inflates the moment you import a 50-word pack you haven't opened once. A graduated count doesn't move until review history backs it up.
One word, one count — even when it changes shape#
Text almost never hands you a word in its dictionary form. An article says despierta; your saved word is despertar. Without collapsing those together, the same verb would quietly count itself several times over as you encountered its conjugations — and your "vocabulary size" would really be measuring how inflected the language is, not how much you know.
LingoBlend resolves this the same way it resolves inflections everywhere else in the app: through the vocabulary graph's form resolution. Every saved word is counted under its single dictionary form — comer, not como/comes/comía/comimos as four separate entries — so mastering a verb across its forms adds exactly one word to your count, not one per conjugation you happened to save. It's the same lemma-collapsing that lets blends skip words you already know recognize despierta as belonging to despertar you've mastered, rather than treating it as an unfamiliar word.
Where the CEFR-ish label comes from#
The count alone ("312 words") doesn't mean much without a reference point, so it's paired with a rough level label — the same one Smart Blend already uses to decide which words are worth putting in front of you. That's not a coincidence: your vocabulary stat and your reading personalization are read from one shared engine, so they can't quietly disagree with each other.
| Self-reported starting point | Label shown |
|---|---|
| Just starting | A1 |
| I know the basics | A2 |
| I can hold conversations | B1 |
| I'm comfortable reading | B2+ |
That starting point is the same one from level-aware blending: what you pick at onboarding, gradually superseded by your measured vocabulary once you've built up enough real review history for your data to outweigh your guess. Improve past your self-assessment and the label moves up on its own; overestimate yourself and it settles back down. Nobody has to remember to update a setting.
Worth saying plainly: the label is a working shorthand for personalizing your blends, not a peer-reviewed conversion table from word count to CEFR band — published academic estimates for what a given CEFR level "means" in raw vocabulary vary considerably, and we cover that properly in how to measure your vocabulary size. Treat the label here as directionally useful, the same way we ask you to treat "roughly" in the count itself.
The honest part: a floor, not a ceiling#
Here's the part worth being upfront about, because it's easy for a number like this to overclaim. LingoBlend's vocabulary count is not an estimate of your total knowledge of a language. It counts words you've learned and proven inside LingoBlend — through saving, reviewing, and graduating them in spaced repetition. If you walked in already knowing 500 Spanish words from years of high school classes, and you've never saved or reviewed them here, they don't count. The number can only grow from what the app has actually watched you learn.
That makes it a floor, not a ceiling. A "roughly 320 words" learner might genuinely know a great deal more than 320 words — this is just the part the app has evidence for. We'd rather show a number that's honestly conservative than one that's flatteringly wrong, because the whole point of the stat is that you can trust it. It's also why vocabulary size is worth tracking at all: research on reading has repeatedly found that how many words you recognize is one of the strongest predictors of how much of a text you actually comprehend (Nation's frequency-and-coverage work is the standard reference here) — so a number that's honest, even a modest one, is more useful than one that's generous but made up.
Where you'll find it#
The stat lives in two places, both read-only and both gated on having enough data to mean something: the Practice screen shows it as a card once you've graduated enough words in the language pair you're currently learning for the number to be meaningful, and the same figure appears again in the detailed all-time stats view alongside your streak and accuracy. Before that, nothing renders — a "3 words" stat helps nobody, and an empty card is more honest than a noisy one.