Every widely used readability formula takes the same two inputs: how long the sentences are, and how long the words are. Nothing else. They cannot see vocabulary, structure, logic or whether the passage means anything at all.
That is worth knowing before treating a score as a verdict, because it is trivially gameable. Splitting every sentence in half improves the score and may make the writing worse.
What do the formulas actually compute?
A weighted combination of those two averages, scaled to a familiar range.
| Formula | Output |
|---|---|
| Flesch Reading Ease | 0 to 100, higher is easier |
| Flesch-Kincaid Grade | a US school grade level |
| Gunning Fog | years of education needed |
The reading ease score subtracts a multiple of the average sentence length and a larger multiple of the average syllables per word from a fixed constant. A result between 60 and 70 is usually described as plain English.
The grade-level versions rearrange the same two averages into a number of school years. They agree with each other closely, because they are measuring the same thing on different scales.
Why is syllable counting the weak point?
Because it cannot be done reliably by rule. Counting vowel groups and discarding a silent final e works for most English words and fails predictably on the rest, as the counting article sets out.
The error mostly cancels out over a long passage, which is why the formulas work at all. On a short one it does not, so a score computed over two paragraphs is far less stable than one computed over two thousand words.
The word-length approach used by one of the formulas — counting words of three syllables or more as complex — inherits the same weakness and adds a threshold effect, since a word either counts fully or not at all.
What can the score not see?
Whether the words are familiar. "The cat sat on the mat" and "The gnu sat on the ait" score identically and are not equally readable, because a formula counting syllables has no notion of which words a reader knows.
It also cannot see order. A paragraph with its sentences shuffled scores exactly the same as the original, which tells you how much of readability these formulas are not measuring.
Jargon is the practical version of this. A page of familiar technical terms scores badly for being polysyllabic while reading easily to its intended audience, and a page of short unfamiliar words scores well while being impenetrable.
How should the number be used?
As a flag rather than a target. A score much worse than expected usually means genuinely long sentences, and long sentences are worth looking at whether or not a formula complained.
Writing to hit a number produces a characteristic style: short, choppy, monotonous sentences that satisfy the arithmetic and are tiring to read. Varying sentence length is what makes prose readable, and a formula measuring the average is blind to variation.
The useful workflow is to check the score, find the longest sentences, and decide about each one individually. That uses the formula for what it can do and leaves the judgement where it belongs.
What actually helps readability?
Shorter sentences do genuinely help, which is why the formulas correlate with anything at all. Beyond that, the improvements are things no formula measures: putting the point first, using concrete nouns, cutting qualifiers, and breaking a wall of text into sections a reader can navigate.
Sentence length variation matters more than sentence length average. A long sentence after several short ones reads fine; five long ones in a row do not, and both sets can share a mean.
The one measurement worth taking alongside the score is the ratio of words to sentences, which is the same input expressed in a form you can act on directly.
Where did the formulas come from?
Wartime and post-war work on making official documents usable, refined for the US Navy in the 1970s to check whether technical manuals could be read by the people issued with them.
That origin explains both their strength and their limits. They were built to flag documents that were unreadable by any measure, not to fine-tune prose that was already clear — and used for the first job they work well.
It also explains the school-grade output. The intended reader was an administrator deciding whether a manual matched a training level, which is a different question from whether an article reads nicely.
Questions people ask
Should headings be counted? Most tools count them as sentences, which shortens the average and flatters the score. Check the passage without them if it matters.
Do search engines use readability? Not as a stated ranking signal. Whether people stay and read is measured differently.
Which formula should I use? Any of them, consistently. Comparing scores from different formulas is meaningless.
Does a low grade level mean simplistic? No. Plain writing about complex subjects scores well and is harder to produce than jargon.
Do the formulas work in other languages? Not without recalibration. The constants are fitted to English.
Is there a target for web writing? Around a grade level of eight to ten suits general audiences, and it is a guide rather than a rule.
Measure the inputs, then read the long sentences. The word counter and sentence counter give the ratio the formulas run on, the syllable counter supplies the other input, and the reading time calculator answers the question people usually meant to ask.