Readability checker: reading ease, grade level and long sentences

What the score actually measures

Two things, and only two: how long your sentences are, and how many syllables your words have. Flesch reading ease combines them into a number from 0 to 100, where higher is easier. That is the whole formula. It does not know whether a sentence is clear, whether a paragraph follows from the one before it, or whether the piece says anything true.

This is worth stating plainly because readability scores get treated as quality scores, and they are not. A page of short sentences made of short words can score 80 and still be unreadable, because it repeats itself and never arrives anywhere. A dense technical paragraph can score 35 and be exactly right for the person reading it.

What the score is genuinely good for is catching drift. Writing gets longer-winded under pressure — more hedging, more subordinate clauses, more abstract nouns — and a score that has slipped ten points since your last article is a reliable signal that something happened to the prose. The sentence list below the score is where the fix is.

• Reading ease 60 to 70 is plain English: most published web writing aims here.
• Below 50 is heavy going for a general reader, which may be correct for a specialist one.
• Above 80 usually means short sentences, which is not the same as clear ones.

What to change, in what order

Start with the long sentences, because they are the highest-value edit and the only part of this report that points at specific text. A sentence over twenty-five words almost always contains two sentences and a conjunction. Splitting it costs nothing and improves both halves — and it moves the score more than any amount of word substitution.

Then look at the syllables-per-word figure. If it is high, the usual cause is not vocabulary but nominalisation: "carry out an assessment of" instead of "assess", "is a requirement for" instead of "needs". Those phrases are long, abstract and slow, and replacing them is the single most reliable way to make a paragraph read faster without dumbing anything down.

Ignore the score itself after that. Once the long sentences are split and the nominalisations are gone, the remaining difficulty is usually the subject, and making the subject sound easier than it is would be a disservice. Write for the reader you actually have.

Why French gets a different formula

Flesch reading ease was fitted on English prose, and French words carry more syllables per word on average. Running the English formula on French text reports a score several points lower than the same prose deserves, which makes the number useless for comparison.

So French uses the Kandel–Moles adaptation: the same shape, with the constants refitted for French. It is still an approximation of sentence and word length, with the same limits as the English version, but at least it is calibrated for the language being measured. The French version of this tool is at /fr/outils/test-de-lisibilite.

Where this fits in a workflow

At the review step, not during drafting. Checking readability while you are still deciding what to say is a way of avoiding the harder question, and editing for a score before the structure is settled means editing sentences you may well cut.

In PrismDraft the review step is a real state: an article waits in Needs review with its outline, its prose, its images and its credit usage visible together, and only an approved article can be published. That is the moment to paste a section in here, split the long sentences, and approve. The score is not stored anywhere and does not need to be — it is a prompt to make an edit, not a metric to report.