What Level Does Duolingo French Get You To?

Two answers exist to this question and they are almost three CEFR levels apart.

Duolingo’s answer is B2. On 22 April 2026 the company announced that nine of its courses, French included, now teach content up to a Duolingo Score of 129 - the top of its B2 band - where most of them previously stopped at A2, a Score of 59.

The research answer is A1. The only peer-reviewed measurement of what Duolingo learners actually reach found Intermediate Low reading and Novice High listening on the ACTFL scale, which ACTFL’s own crosswalk converts to A1.2 and A1.1.

Neither number is dishonest. They measure different things, and the distance between them is the whole subject of this article.

What does Duolingo itself claim?

That the course now goes to B2, which is a real change and worth taking seriously.

The April 2026 announcement extended nine courses - English, Spanish, French, German, Italian, Portuguese, Japanese, Korean and Chinese - from a ceiling of A2 to a ceiling of B2, expressed in the company’s internal Duolingo Score as a jump from 59 to 129. Before that, a learner who finished the entire French tree had, by Duolingo’s own alignment, completed A2 material.

That is a claim about content, and content claims are cheap to verify and easy to trust. If Duolingo says its French course contains subjunctive triggers and the sequence of tenses, it does.

It is not a claim about outcomes. The distinction is the same one that separates a syllabus from a transcript, and no course anywhere converts one into the other at a rate of 100%.

What has actually been measured?

One study, published in a serious journal, and it is more specific than most people realise.

Jiang, Rollinson, Plonsky, Gustafson and Pajak, Evaluating the reading and listening outcomes of beginning-level Duolingo courses, Foreign Language Annals 54(4), 974–1002, 2021 - the ERIC record is public. Its design:

• 225 participants, US-based, with little or no prior proficiency.
• Learning Spanish or French, with Duolingo as their only tool.
• Assessed after completing the then beginning-level courses.

The findings, verbatim: learners “reached Intermediate Low in reading and Novice High in listening”, and their scores “were comparable with those of university students at the end of the fourth semester of study”. Those who completed seven units matched roughly five university semesters.

Two caveats belong next to that, in fairness and in both directions. The study is not independent - Duolingo promoted it through its own investor communications - and it was still peer-reviewed in the field’s leading journal, which is more than any competing app can show. And no other skills were assessed: no speaking, no writing.

What is Intermediate Low in CEFR terms?

A1 - and this is the step almost every article about Duolingo gets wrong.

ACTFL publishes an official crosswalk to the CEFR, and the crucial detail is that it has two different columns. The same ACTFL rating maps to a lower CEFR level for receptive skills than for productive ones, because reading a sentence is easier than producing it.

The Duolingo study measured reading and listening. So the correct column is the left one, and the published outcome reads A1.2 in reading, A1.1 in listening.

Quote the right-hand column instead - as almost everyone does, usually by accident - and the same result becomes “A2 reading, A1 speaking”, which sounds a level better and describes skills the study never tested.

Why is the gap between the claim and the measurement so wide?

Four reasons, and only one of them is Duolingo’s fault.

The study predates the expansion by five years. It measured the old beginner courses. Nobody has measured outcomes from the new B2 content, because it shipped in April 2026. That cuts in Duolingo’s favour and should be said plainly.

Almost nobody finishes. Course ceilings describe the last unit, and completion rates for any self-directed course are low. The level of the final lesson is not the level of the median user.

Recognition outruns production. Tapping the right tile among four is a different act from building the sentence. The whole apparatus measures the easier one, which is exactly what ACTFL’s two-column crosswalk formalises. It is also the gap that no app closes on its own, since most French agreement errors are inaudible and therefore invisible to any exercise that never asks you to write.

French punishes this pattern specifically. Written French and spoken French have drifted far apart - qu’est-ce qu’il y a is five written words and about two syllables heard - so an app-built vocabulary is systematically over-strong on the page and under-strong in the ear. The study found precisely that shape: reading a full sublevel above listening. That asymmetry is structural to French, not an artefact of the app.

What does Duolingo genuinely do well in French?

More than the sneering suggests, and the honest case for it is a strong one.

It is very good at the A1 block, which is not a small thing: in our own curriculum A1 holds 99 grammar points, more than any other grade and more than C1’s 62. Getting a beginner through that mass of articles, gender, agreement, negation and the present tense is genuinely hard, and drilling is the correct method for material that is finite and mechanical.

It is also good at habit, which is the constraint that actually stops most people, and at early vocabulary, where spaced repetition works.

Matched against the roadmap, that puts it exactly where the complete roadmap for learning French puts apps in general: the right tool for the first stage, and the wrong one for the stages after it.

Where does it stop paying?

Around A2, which is also where the standard plateau begins - and the two facts are related.

At A2 the bottleneck stops being coverage and becomes volume of real input: text you understand at around 98%, in quantity. An exercise interface cannot deliver volume, because every item is short by design. That is the argument for switching to graded material, set out in how to learn French by reading.

From B1 the bottleneck moves again, to precision under correction. B2 is where French stops rewarding you for getting the message across and starts judging how you got it across - the shift described in B1 vs B2 French. No app corrects the subjunctive you never attempted.

If your target is administrative rather than personal, this matters more than it sounds: B2 across all four skills is now the French naturalisation threshold, as covered in the guide to French levels for citizenship.

So what level does it get you to?

The defensible answer, stated with its conditions:

• Course content: through B2, since April 2026, on Duolingo’s own alignment.
• Measured outcome, reading and listening: around A1, from the beginner courses, in the only peer-reviewed study.
• Measured outcome, speaking and writing: unknown, because it has never been tested.
• Reasonable expectation today: a solid A1 and a workable A2 if you finish a lot of it, with listening trailing reading, and production untested.

Use it for the first stage and stop expecting it to do the later ones. What comes next is a level-by-level sequence rather than another app - our A1 to C1 curriculum publishes all 61 levels and 424 grammar points unit by unit, so you can see exactly which structures you have and which ones you have been quietly avoiding.