How to Read French News: The Real Difficulty

The easiest French news to read is RFI’s Journal en français facile, but not for the reason you have been told. Its vocabulary is barely lighter than a national outlet’s - what it actually cuts is sentence length, by a third.

We collected 73 articles across four outlets in July 2026 and ran them through the same frequency pipeline we used on eleven French classics: the same reference corpus, the same lemma ranking, the same denominator. That matters, because the two tables are only worth putting side by side if nothing changed between them.

How much of a French news article can you read?

Each column is the share of running text covered by the N most frequent headwords of written French.

Two things stand out immediately. The spread between the easiest and hardest source is small - 3 points at the widest vocabulary - and every one of these figures is worse than almost every novel we measured.

Why is the press harder than the classics?

Because news is made of names, and a frequency list built on books does not contain them.

Run through the identical pipeline, the classics land between 96.8% and 98.7% at 11,338 headwords. The press lands between 93.0% and 96.0%. On the headline number the nineteenth century wins, which is not the result anybody expects.

The explanation is in the unknown words rather than the known ones. Between 3.9% and 8.1% of press tokens are absent from the reference corpus entirely, and between 40% and 73% of those absences are proper nouns: Zelensky, Bercy, Décathlon, Tarn-et-Garonne. You do not learn those. You see the capital letter, register that it names a person or a place, and read on. Counting names as already acquired, the real ceiling on raw press text is:

Then there is the tense that decides everything. Across all 61,627 words of press we found seven unambiguous passé simple verbs. Seven. The same measurement on novels returns 95 to 315 per 10,000 words - roughly 150 times the density. The press asks you to absorb names and topical nouns, which is cheap and self-correcting. A novel asks you to hold a tense system you have never conjugated, which is neither.

What makes “français facile” easy, then?

Sentence length, almost entirely.

RFI’s simplified bulletin reaches 96.0% coverage where franceinfo reaches 94.4%. That is a real gap, but it is 1.6 points - nothing like the difference in reputation. Meanwhile RFI’s sentences average 15.0 words against franceinfo’s 23.5, a third shorter, and 20 Minutes sits at 22.7 despite its reputation for plain writing.

RFI also has the highest proper-noun density of the four, at 5.4% of text, because a world-news bulletin is wall-to-wall countries and ministers. So “français facile” is not a smaller word list. It is a shorter sentence, one clause at a time, in a register that has stopped hiding the verb behind three subordinate clauses.

That is worth knowing because it changes how you use it. If your problem is that you know the words but lose the thread halfway through the sentence, RFI is exactly the right tool. If your problem is vocabulary, it will barely help.

Why is the children’s paper the hardest of the four?

For the same reason children’s classics are, and this is the most useful finding here.

1jour1actu is written for French nine-year-olds. It has the shortest sentences in the study at 14.0 words, and the smallest vocabulary at 1,249 distinct headwords. Yet it finishes last on coverage - 93.0% at 11,338 headwords - and would need around 27,000 headwords to reach 98%, the worst figure we have measured on any text.

Its proper-noun rate is 1.6%, the lowest of the four, which is the tell. When a word is unknown in 1jour1actu, it is usually a real French word: school vocabulary, French institutions, the ordinary furniture of a French childhood. Native children arrive with all of it. Learners do not, and none of it appears in a frequency list built on adult prose.

This is precisely the pattern we found in Les malheurs de Sophie - small vocabulary, unusable tail - and it confirms something we had previously only asserted in French graded readers by level: material written for native children is not the same thing as material written for learners, and the difference runs the wrong way.

Two caveats on this row. It is the smallest corpus in the study at 8,885 words, so its figures are the least stable of the four. And we removed two items by hand before measuring - a reader’s letter written by a ten-year-old and a competition announcement - because neither is editorial prose, and leaving them in would have made the source look easier than it is.

Does the difficulty stop growing if you keep reading?

Yes for some outlets, and no for others, which is the practical finding.

We counted how many new headwords each successive 2,000-word block of a single source introduced:

RFI’s cost falls by 76% between the first block and the seventh: it recycles a controlled vocabulary, so the entry fee is paid once. 20 Minutes barely falls at all, because a general daily covers whatever happened that day and each new subject imports its own nouns.

The operational consequence: one outlet, read daily, is a real strategy. Hopping between outlets is not - you keep paying the first-block price. If you narrow further and follow one beat, sport or politics or tech, the curve drops faster still.

Which outlet fits your level?

A2 to early B1 - RFI Journal en français facile. Short sentences, controlled vocabulary that repeats, and a transcript beside the audio. The single best entry point in French media.

B1 - 20 Minutes, one section at a time. Longer sentences than RFI, but plain register and short pieces. Pick one section and stay in it so the vocabulary compounds.

B2 - franceinfo or any national daily. Full-length sentences at 23.5 words and no lexical concessions. Reading a general-news article on an unfamiliar subject without a dictionary is close to the definition of B2.

Skip 1jour1actu unless you specifically want French school and childhood vocabulary - a legitimate goal if you are moving to France with children, and a poor one otherwise.

How should you actually read the news?

Four rules, all of which fall out of the numbers above.

Do not look up proper nouns. They are the majority of what you will not recognise, and they cost nothing to skip.

Stay with one outlet for a month. The saturation table is the whole argument: the second month is much cheaper than the first, but only if it is the same source.

Read the same story in two outlets. RFI first, then a full-register version of the same event. The content is known, so the only new variable is the language.

Judge with one page, not one article. If you stop more than twice on a page, the source is above your level today.

That principle - meeting text where the unknown share is small enough to learn from - is why our own material is graded across 61 levels from A1 to C1 rather than curated by taste, so that the line moves with you instead of you guessing where to jump.

Two limits worth stating. Our corpus stops at the free generalist press: Le Monde, Le Figaro, Libération and France 24 all block automated reading, so the upper end of the register is represented by the classics rather than by broadsheet journalism. And this is one snapshot of 73 articles from a single week - a fortnight dominated by different stories would shift the topical vocabulary, though not the structural findings about names, tense and sentence length.