Free Readability Checker

The formulas are easy. The counting is not.

Every readability formula in use today is arithmetic over three numbers: how many sentences you wrote, how many words, and how many syllables. Flesch Reading Ease, for instance, is nothing more than 206.835 − 1.015 × (words per sentence) − 84.6 × (syllables per word). Any two tools given the same three counts will agree exactly.

They do not agree, and the reason is that they disagree about the counts. Specifically, they disagree about syllables. There is no rule in English that reliably tells you how many syllables a written word has: “business” looks like three and is two, “every” looks like two and is three, “chocolate” can be either. The usual approach is to count groups of vowels and patch the result with a handful of rules, which gets roughly 85 to 90% of words right. Reading ease multiplies syllables per word by 84.6, so a counting error of a tenth of a syllable moves the score by eight points and the grade by more than one.

This checker counts syllables by rule and then by exception, with the exception table built from the Carnegie Mellon Pronouncing Dictionary, which records how 126,000 English words are actually said. Measured back against that dictionary and weighted by how often each word is really written, it gets 99.7% of words right, and the total syllable count of a page lands within 0.13% of the dictionary’s own. That is the whole reason to use this one.

What the score means

The big number is Flesch Reading Ease, on a scale where high is easy. Rudolf Flesch published the bands in 1948 and they have not needed changing:

• 90 to 100 — very easy, readable by an average 11 year old.
• 80 to 90 — easy. Most fiction sits here.
• 70 to 80 — fairly easy.
• 60 to 70 — plain English, readable by a 13 to 15 year old. This is the target for most writing on the web.
• 50 to 60 — fairly hard. Quality journalism and good business writing.
• 30 to 50 — hard. University level.
• Below 30 — very hard. Academic papers, contracts, and the kind of policy document nobody finishes.

Scores can go above 100 and below zero. A page of four-word sentences made of one-syllable words will score in the 110s; a dense legal clause can score below −50. Both are real answers, not errors.

Seven formulas and one honest answer

Seven published formulas are reported, because each was built for a different kind of writing and each keys on something slightly different.

• Flesch Reading Ease — the best known, and the one Microsoft Word reports. Sentence length and syllables per word.
• Flesch-Kincaid Grade Level — the same measurements rearranged into a US school grade. Commissioned by the US Navy for technical manuals.
• Gunning Fog Index — sentence length plus the share of words with three or more syllables. Written for business writing, and the strictest on jargon.
• SMOG Index — counts only the long words. The standard choice for health and safety material, because it was validated for complete comprehension rather than partial.
• Coleman-Liau Index — uses letters per word instead of syllables, so it never depends on anyone guessing a pronunciation.
• Automated Readability Index — characters and words, designed in 1967 to be computed by a machine that counted keystrokes.
• New Dale-Chall — the only one that looks at which words you used. A word is hard if it is not among the 3,000 that four out of five fourth-graders know. It is the best validated of the set, because familiarity predicts comprehension better than length does.

Six of those produce a grade level, and the headline figure is their median rather than their mean, so a single formula misbehaving on a short text cannot drag the answer with it. When they agree closely you can trust the number. When they spread out by more than two grades, something unusual is happening — most often short sentences built from long words — and you should read the text yourself.

Three things this does that other checkers do not

It reads acronyms the way you do. “SEO” is three syllables when anyone says it aloud, and “HTML” is four. Tools that treat them as one-syllable words quietly flatter every technical page. Here they are counted letter by letter, as long as the surrounding text is not also in capitals.

It applies Gunning’s own exclusions. Robert Gunning was explicit that proper nouns do not count as hard words, and neither do verbs that only reach three syllables by gaining an ending — “recorded” is not a difficult word. Almost every implementation skips both rules and reports a Fog score that is too high, especially on anything full of names. This one applies them, and spots proper nouns by looking at how a word is capitalised across the whole text rather than guessing word by word.

It shows its working. Every formula is printed with your own numbers substituted into it. You can check any score on a calculator in ten seconds, which is the only real answer to “why should I believe this”.

What to do with a bad score

There are exactly two levers, because there are only two inputs. Shorten sentences, or use shorter and more familiar words. Everything else is decoration.

1. Start with the longest sentences. The hardest sentences are listed worst first. Most of them are two sentences that have been joined with “and” or a comma, and splitting them costs nothing.
2. Then the words you lean on. The hard words list is ordered by how much each one is costing you, which is its length multiplied by how often you used it. A four-syllable word you used once is not the problem; the one you used nine times is. Where a genuine shorter synonym exists it is shown; where it does not, there is a dash, because swapping a precise word for a vague one makes writing worse, not better.
3. Check the marked-up text. Difficulty is rarely spread evenly. Usually two or three paragraphs are carrying it, and the shading shows you where without reading the whole thing again.

What not to do is write to hit a number. Chopping every sentence to nine words produces prose that scores beautifully and reads like a ransom note. The score is a smoke alarm, not a recipe.

Does readability affect SEO?

Not directly. Google does not grade your prose and rank you on it, and has said so plainly more than once. What readability affects is everything that happens after someone lands: whether they read past the first screen, whether they find the answer they came for, whether they come back, and whether anybody thinks the page is worth linking to. Those are not ranking factors either, but they feed the ones that are.

The practical case is simpler than the SEO case. People read on phones, in a hurry, with something else competing for their attention. Writing that survives those conditions wins, and writing that needs a quiet room does not.

How much text does it need?

Thirty words before it will report anything, and a few hundred before the number is worth much. These formulas were developed on passages of a hundred words and up, and SMOG was built around samples of thirty sentences; below that it is extrapolating. The tool says so when your text is short. Run it on the whole article rather than the introduction.

Which tool do I want?

• This one answers “how hard is this to read, and for whom”.
• The Sentence Length Checker goes much deeper on one of the two inputs: the rhythm of your sentences, runs of long ones, and every sentence listed in order.
• The Word Counter is for while you are drafting, with counts updating as you type.
• The Paragraph Length Checker finds the walls of text, which no readability formula measures at all — a page can score 70 and still be unreadable because it has no paragraph breaks.
• The Reading Time Calculator answers how long it takes rather than how hard it is, and adjusts its estimate using the same two measurements this page is built on.
• When the writing is finished, the Blog Title Analyzer, the Meta Description Checker and the SERP Snippet Preview Tool handle how it presents itself in the results.

Is anything stored?

No. Pasted text is measured entirely in your browser and never leaves the page — the pronouncing dictionary and the familiar word list are part of the page itself, which is why the score appears as you type. The one exception is checking a page by address: a browser is not permitted to fetch another site’s pages directly, so the address is sent to our server, which reads the page and returns the text. Nothing is written to a database or a log either way, and the report exists only in the open tab.