The seven formulas behind every readability checker: what each one counts, its exact published coefficients, where it came from, and how far to trust it.
The formulas and their coefficients
Every coefficient below is taken from the original publication. Each formula has its own page with a calculator and an interactive lab.
Kincaid, J. P., Fishburne, R. P., Rogers, R. L., & Chissom, B. S. (1975). Derivation of new readability formulas for Navy enlisted personnel. Research Branch Report 8-75, Naval Technical Training Command.
Senter, R. J., & Smith, E. A. (1967). Automated Readability Index. AMRL-TR-66-220, Aerospace Medical Research Laboratories, Wright-Patterson Air Force Base.
Raw = 0.1579 × (% difficult words) + 0.0496 × (words ÷ sentences), + 3.6365 if difficult words > 5%
Raw score → grade band
Dale, E., & Chall, J. S. (1948). “A formula for predicting readability.” Educational Research Bulletin, 27(1), 11–20; Chall, J. S., & Dale, E. (1995). Readability Revisited: The New Dale-Chall Readability Formula. Brookline Books.
Why formulas disagree: the same passage, six grades
Pick a passage. Each bar is one formula’s grade for exactly the same text. Letter-based formulas (Coleman-Liau, ARI) react to long words differently from syllable-based ones, and SMOG aims for full comprehension, so it usually sits higher.
Most people check their email before breakfast. That habit feels useful, but it hands your first hour to other people. Try a small change this week. Leave your phone in another room until you have eaten and planned your day. You will still reply to everyone. You will just do it on your terms, and the work that matters will get your best attention.
Flesch-Kincaid3.7
Gunning Fog6.1
SMOG7.2
Coleman-Liau7.0
ARI4.4
Dale-Chall (band)4.0
Spread: 3.5 grades between the highest and lowest formula. Average: 5.4. This is why the readability checker shows all of them and an average.
How we count words, sentences and syllables
The formulas are exact; the counting that feeds them is where every tool makes choices. Ours, so you can compare:
Words
A word is a run of letters or digits. Contractions (“don’t”) and hyphenated compounds (“well-known”) count as one word. Markdown heading lines (starting with #) are excluded from readability statistics but used for the heading checks; list markers are ignored.
Sentences
A sentence ends at “.”, “!”, “?” or “…” followed by a space, closing quote or end of text, and also at a line break, so headings, list items and captions don’t merge into one giant sentence. Common abbreviations (Mr., Dr., e.g., i.e., etc., month names), single initials and decimal numbers do not end a sentence. Unusual abbreviations can still split a sentence early, which slightly lowers grade scores.
Syllables (the approximate part)
Syllable counts come from the MIT-licensed syllable library by Titus Wormer. It is based on the rules in the PHP Text-Statistics project, itself inspired by the Perl module Lingua::EN::Syllable. It counts vowel groups (y counts as a vowel except at the start of a word), subtracts silent endings such as final “e”, “-ed” and “-es”, adds splits like “-ia” and “-io”, and overrides the result for a list of irregular words.
This is a heuristic, not a pronouncing dictionary. It is right for the large majority of common words but can be off by one on unusual words, names and words with two accepted pronunciations (“fire”, “every”). Numbers written as digits are counted as if each digit were read aloud. On a typical 500-word passage the error is usually well under one Flesch point and a fraction of a grade. See the rules applied to any word on the syllable counter.
Letters and characters
Coleman-Liau uses letters only (A–Z including accented letters). ARI uses letters and digits. Neither counts punctuation or spaces.
Complex words, polysyllables and difficult words
SMOG polysyllables: every word of three or more syllables, repeats included.
Gunning Fog complex words: three or more syllables, excluding capitalised words that aren’t sentence-initial (proper nouns), hyphenated compounds, and words that only reach three syllables through -es, -ed or -ing. Gunning also excluded familiar jargon, which software can’t detect, so it isn’t applied.
Dale-Chall difficult words: any word not on the 1995 list of 2,942 familiar words (from the MIT-licensed dale-chall dataset). Regular inflections of listed words (-s, -es, -ed, -ing, -er, -est), proper names and numbers count as familiar, following Chall and Dale’s counting rules. Irregular forms not on the list are counted as difficult.
Test passages and results
Two kinds of test run on every build (scripts/test-formulas.ts in the source):
Coefficient checks. Each function is compared with hand calculation from the published equation. Example: “The cat sat on the mat.” has 6 words, 1 sentence and 6 syllables, so Flesch Reading Ease = 206.835 − 1.015 × 6 − 84.6 × 1 = 116.1, which the checker returns.
A published worked example. “The Australian platypus is seemingly a hybrid of a mammal and reptilian creature.” is widely used to illustrate Flesch-Kincaid: 13 words, 26 syllables, 1 sentence, grade 13.1. The checker counts 13 words and 26 syllables and returns 13.1.
And the four sample passages used across the site, from picture-book prose to an academic abstract, scored in plain code:
Passage
Words
Sent.
Syll.
FRE
FKGL
Fog
SMOG
CLI
ARI
Dale-Chall
Picture-book prose
31
5
31
115.9
-1.4
2.5
3.1
-2.7
-4.0
0.31
Plain blog intro
64
6
82
87.6
3.7
6.1
7.2
7.0
4.4
1.02
Corporate policy
51
3
109
8.8
16.3
18.6
16.7
19.1
16.4
11.91
Academic abstract
45
3
159
-107.3
32.0
34.4
22.1
36.8
29.8
15.96
The formulas rank all four passages in the same order. They disagree on absolute grade, most of all on the academic abstract, where long Latinate words push letter-based and syllable-based formulas far apart. SMOG values on these short passages are rough because SMOG expects 30 sentences. Expect other tools to differ by a point or two on the same text because of the counting choices above.
How to improve readability
Split long sentences. Anything over 25 words is a candidate. Break at “and”, “but”, “which” or a semicolon.
Prefer everyday words. “Use” for “utilise”, “help” for “facilitate”, “about” for “approximately”. Each swap cuts syllables, the heaviest input in Flesch.
Put the actor first. Active voice is usually shorter and clearer. The passive voice checker finds candidates.
Explain unavoidable jargon the first time, in a few plain words.
Break up the page. Subheadings every 250–300 words, paragraphs of two to four sentences, lists for steps.
Read it aloud. Wherever you stumble or run out of breath, a reader will too.
Writing for a lower grade is a skill, not a downgrade
Grade 8 targets make some writers worry about sounding childish. The opposite is usually true: the text gets shorter and more confident. Explaining why the Moon has phases to a ten-year-old takes real precision, and ahaboo is a good study in how plain words and short steps can carry a hard idea.
If you want a first draft at a lower grade to edit from, the AI rewrite can simplify a passage to a target grade and re-score the result.
Readability formulas FAQ
Which readability formula is the most accurate?
None is best everywhere. Each was validated on different readers and texts. Flesch-Kincaid is the most widely used, SMOG is favoured in health care for predicting full comprehension, and Dale-Chall is the only one that checks vocabulary. Averaging several reduces each one’s blind spots.
Why do readability tools give different results?
The equations are fixed, but counting isn’t. Tools differ in how they split sentences (headings, lists, abbreviations), which characters they count, and how they estimate syllables. A grade apart between tools is common.
How do I improve readability?
Shorten long sentences, replace long or rare words with everyday ones, prefer active voice, use subheadings and short paragraphs, and explain necessary jargon the first time you use it.
Are readability formulas used by Google?
No. Google has said readability scores are not a ranking factor. They are editing tools for writers.