French exercises online: what free practice fixes, and what it never will

The règlement des épreuves annexed to the Bulletin officiel sets the same four papers at A1, A2, B1 and B2: compréhension de l'oral, compréhension des écrits, production écrite, production orale. Each is marked out of 25. The total is out of 100. That is the whole architecture.

Grammar is not one of them. Nor is vocabulary, nor conjugation, nor spelling. They are everywhere in the papers — you cannot write a B1 essay in the wrong tense and score — but none of them is examined as an object. The single most searched category of free French exercise trains something that does not appear on the mark sheet under its own name.

This is not an argument for skipping grammar. It is an argument for knowing what a grammar exercise is: a means, whose only evidence of having worked is a sentence you produce later, in one of the four papers. A page of correct answers is not that evidence. It is a step.

The règlement describes each paper in a line or two, and those lines are a better shopping list than any exercise index.

Now the piles sort themselves, and the sorting is not flattering to the biggest one.

Multiple-choice grammar and conjugation drills — the "with answers" category — train recognition and, when the answer is typed rather than chosen, recall. They map onto none of the four papers directly. Their value is entirely indirect and entirely real: they make the tense available when the sentence needs it.

Gap-fill and cloze exercises sit one step closer. They ask for production inside a frame someone else built. Nothing in the DELF looks like them, which is worth knowing before spending a term on them; they remain the cheapest way to find out that the subjunctive is not yet automatic.

Reading-comprehension exercises with questions are the closest free match to an actual paper — the format is the same, and half the exam is comprehension. This is where free material earns most.

Listening exercises come next, with one condition attached, and it is the condition that most free practice quietly removes: the number of times you may play the recording. Two listenings at A1, A2 and B1. At B2, the interview or bulletin is played once. A learner who habitually rewinds has been practising a different exercise.

Writing prompts and speaking prompts are the two halves of the exam that carry fifty marks out of a hundred, and they are the two for which no free exercise can close the loop. Not because the prompts are hard to find — the sample papers publish them — but because nothing marks what comes out.

The completion "with answers" appears on almost every root in this cluster, and it is a reasonable thing to want. It is also the point at which the free supply and the official supply part company.

The official sample papers for the DELF are distributed for every level, free, with the audio files. What they do not include, at any level, is a corrigé, a marking grid or an answer key of any kind. Not for the comprehension papers, and certainly not for the two production papers, which have no key to publish.

So the answer keys that circulate for French exercises come from publishers and from teachers, not from the body that sets the examination. That is fine for a conjugation table. It is worth remembering when a key claims to correct a piece of writing: what it is offering is a model answer, which shows what a good response looks like without saying anything about the one you produced.

The distinction is small in words and large in practice. A key marks a choice. A correction reads a sentence that did not exist before you wrote it.

This site carries a free exercise bank: one tab per level from A1 to C2, 149 questions and 6 written-production tasks in total, 24 to 25 questions in a level's session. A session is built like a paper rather than like a worksheet — sixteen sentence completions, a set of documents with questions on them, and one writing task with a model.

The answers do not appear until the whole set has been submitted. That is deliberate, and it is the one interface decision in the bank worth defending in writing.

Correcting question by question feels better and destroys the measurement. Once the previous answer has been confirmed, the next one is no longer attempted from what you know; it is attempted from what you have just been told, from the pattern of the last three answers, and from the shape of the options. The score at the end stops describing your French and starts describing your ability to read a quiz. Two things are lost at once: the number, and the honest experience of not knowing — which is the experience the examination room supplies for free.

The trade-off is real and goes the other way too. Delayed feedback is less comfortable, and on a wrong answer you carry the mistake for another twenty questions. That is the cost of getting a number that means something. Anyone who wants immediate confirmation is choosing practice over measurement, which is a legitimate choice on a Tuesday and not on a mock.

If the four papers are the yardstick, an hour splits fairly obviously — and not the way free material's availability would split it.

Fifteen minutes of drilling, on the same material rather than on new material. Repeating one set until it is automatic does more than covering three new ones, and the free supply is optimised for the opposite, because new sets are what a site has to publish to stay visited.

Twenty-five minutes of comprehension, timed, with the audio played the number of times the level allows and no more. This is the half of the examination that free material genuinely serves.

Twenty minutes of production, written or spoken, from a real prompt. Write the text. Say the presentation out loud and record it. You cannot mark either, and you will still learn more in those twenty minutes than in an hour of gap-fills, because you will discover which structures you avoid — and avoidance is invisible in any exercise where the structure is given to you.

The last one is the point at which the honest version of this guide has to stop being about exercises.

Free practice takes a learner a long way on two of the four papers and a measurable distance on the vocabulary and grammar that feed the other two. It cannot do three things, and none of them is a matter of finding a better site.

It cannot mark a sentence you invented. It cannot hear that a vowel is wrong, because you do not hear yourself being wrong. And it cannot tell you what you are avoiding — the construction you never attempt, which never produces an error and therefore never produces a correction.

That is a description of a limit, not a sales argument, and it does not disqualify anything. A learner who drills alone for six months and then has someone read one text has spent those six months well. A learner who drills alone for six months and never writes anything has been practising the smaller half.

Our lessons are individual and by video, with teachers based in France. General French is 45 €/h and examination preparation is 50 €/h, invoiced in euros with no platform commission; hours are not sold as a subscription and unused hours are refunded in money rather than in credit. A Singapore evening falls in the early afternoon in France, which is a wider slot than the one local tutors are all asked for.

The free bank on this site is a fixed snapshot — it is written once and nothing feeds it automatically. The students we teach work from the same bank with new items added every day, and the items they get wrong are taken apart in the lesson rather than shown as a corrected answer. The gap between those two versions is the whole of what a lesson adds, and it is easier to see than to describe.

The first twenty minutes are free and by video. Bring the exercises you are already doing: the useful question is not whether they are good, but which of the four papers they are training, and which one nobody is training.