Text Readability Calculator

Introduction to readability scoring for English prose

Readability formulas are regression equations. Each one was fitted, decades ago, against the comprehension scores of real readers on a specific set of graded passages, and each one predicts difficulty from two or three surface features of the text that a machine can count without understanding a word of it: how long the sentences are, how long the words are, and how many of those words are polysyllabic. Nothing in a readability formula knows what your sentences mean. What the formulas do offer is a fast, reproducible, comparable number — and that is genuinely useful when you are editing a draft, standardising a document set, or checking whether public-facing material is within reach of the audience that has to act on it.

This calculator reports six published indices side by side rather than one. Reporting only Flesch Reading Ease, as most quick tools do, hides the fact that the indices disagree, sometimes by three or four grade levels on the same passage, because they weight vocabulary and sentence length differently and were calibrated on different corpora. Seeing the spread is the point: a passage where Flesch-Kincaid says grade 7 and Gunning Fog says grade 13 is a passage with short sentences stuffed with long words, and that diagnosis is more actionable than any single score.

The six indices are the Flesch Reading Ease score (Flesch, 1948), the Flesch-Kincaid Grade Level (Kincaid and colleagues, 1975, for the US Navy), the Gunning Fog Index (Gunning, 1952), the SMOG grade (McLaughlin, 1969), the Coleman-Liau Index (Coleman and Liau, 1975) and the Automated Readability Index. Five of them return a US school grade level. Flesch Reading Ease alone runs the other way, on a 0-to-100 scale where higher means easier.

Everything runs in your browser. The passage you paste is never sent anywhere, which matters if you are checking unpublished copy, patient material, internal policy, or anything else you would not paste into a third-party service. The page also reports the raw counts it derived — words, sentences, syllables, letters, complex words, polysyllables — so you can audit any score by hand against the equations below rather than trusting the output.

How to use this readability analyser on a working draft

1. Paste the passage into the text box. Use a representative sample of at least 100 words; a single sentence will produce a number, but it will be a number with no stability behind it.
2. Pick the reading grade your audience actually has. The default of grade 8 reflects the common target for general public-facing material; health and safety communication is usually pitched lower, at grade 6 or below.
3. Select Analyse readability. The status line gives a one-sentence verdict; the panel underneath breaks out all six indices, the counts they were computed from, and the sentences that are dragging the score.
4. Work on the flagged sentences first, then re-run. The consensus grade is the median of the five grade-level indices, which is more robust to a single outlier than an average.
5. Use Copy summary or Download CSV to record the result alongside the draft, so that a later revision can be compared against the same baseline.

Two habits make readability scores worth having. First, always compare a passage against itself before and after editing rather than against an absolute threshold — the formulas are far better at ranking two versions of the same content than at pronouncing on one version in isolation. Second, sample the same way every time. Scores computed on 80 words of introduction are not comparable with scores computed on 800 words of body copy, because sentence-length variance shrinks as the sample grows.

Every readability formula on this page, written out

Let be the number of words, the number of sentences and the number of syllables in the sample. Two ratios do most of the work in every formula below — average sentence length and average syllables per word:

Formula: ASL = W / S, ASW = Y / W

Flesch Reading Ease

Flesch published this in 1948 as Formula A, written in terms of word length measured as syllables per 100 words and sentence length measured in words: . Because is per 100 words, the modern per-word restatement multiplies that coefficient by 100, giving the identical equation used here:

Formula: RE = 206.835 - 1.015 ⁢ W / S - 84.6 ⁢ Y / W

Flesch scaled the result so that 100 corresponds to material a reader who has finished fourth grade can handle, and he noted in the original paper that the scale under-rates grade level beyond about seventh grade. There is no theoretical lower bound: one very long sentence of very long words can drive the score below zero.

Flesch-Kincaid Grade Level

Kincaid, Fishburne, Rogers and Chissom re-fitted Flesch's variables against the comprehension of 531 US Navy enlisted personnel and rescaled the output to a grade level, so that higher now means harder:

Formula: FKGL = 0.39 ⁢ W / S + 11.8 ⁢ Y / W - 15.59

Gunning Fog Index

Gunning's index adds a vocabulary term that counts hard words — words of three or more syllables, excluding proper nouns, familiar jargon and compounds, and not counting a common inflectional ending such as -es, -ed or -ing as one of those syllables. With the count of hard words:

Formula: Fog = 0.4 ⁢ W / S + 100 ⁢ C / W

SMOG grade

McLaughlin's Simple Measure of Gobbledygook counts polysyllables — words of three or more syllables, this time including proper nouns and repetitions — in a sample of 30 sentences. The published regression, with the polysyllable count, normalises any sample size back to 30 sentences:

Formula: SMOG = 1.0430 ⁢ sqrt(P ⁢ 30 / S) + 3.1291

The hand-worked version McLaughlin promoted for practitioners rounds this to "take the square root of the polysyllable count in 30 sentences and add 3", which is the form reproduced in health-literacy handouts. This page uses the full regression constants, so it will differ from a pencil-and-paper SMOG by a few tenths of a grade.

Coleman-Liau Index

Coleman and Liau argued that letters per word predict difficulty better than syllables per word, and crucially that letters can be counted mechanically without a pronunciation dictionary. Their index takes as letters per 100 words and as sentences per 100 words:

Formula: CLI = 0.0588 ⁢ L - 0.296 ⁢ S ′ - 15.8

Automated Readability Index

The ARI likewise avoids syllables, using characters per word. It was designed for real-time scoring from a typewriter keystroke counter, which is why it needs nothing but characters, words and sentences:

Formula: ARI = 4.71 ⁢ Ch / W + 0.5 ⁢ W / S - 21.43

How this page counts words, sentences and syllables

Every readability score is only as good as its counters, and counters are where implementations quietly diverge. These are the rules this page applies, stated so you can reproduce them.

• Words. A word is a run of letters, optionally joined by an apostrophe or hyphen (doesn't, state-of-the-art), or a run of digits with internal separators (3.14, 1,200). Punctuation is stripped before counting, so a trailing full stop never inflates a word's letter or syllable count.
• Sentences. A sentence ends at a full stop, exclamation mark or question mark that is followed by whitespace or the end of the text. A lone full stop is not treated as a terminator when it sits between two digits (3.14), when it follows a single letter (J. P. Kincaid), or when it follows one of about seventy recognised abbreviations (Dr., Fig., etc., vs., Inc., month and weekday abbreviations). Closing quotation marks and brackets are kept with the sentence they close.
• Syllables. Counted by a vowel-group heuristic with corrections: a final silent e is removed (make = 1), but a consonant plus -le is restored as its own syllable (table = 2, candles = 2); a regular past-tense -ed is silent unless it follows t or d (walked = 1, wanted = 2); a plural -es is silent unless it follows a sibilant (makes = 1, boxes = 2); a word-initial y is a consonant. A small exception list catches common irregulars the heuristic gets wrong (area, business, science, every). Digits are read digit by digit.
• Complex words versus polysyllables. These are deliberately different counts. Gunning's complex words exclude proper nouns and hyphenated compounds and ignore inflectional endings; SMOG's polysyllables include everything with three or more syllables. Tools that use one number for both will disagree with this page.
• No terminal punctuation. If the passage contains words but no sentence-ending punctuation at all, the whole passage is treated as one sentence and the result is flagged, rather than dividing by zero.

Worked example: a patient appointment reminder

Take a three-sentence reminder of the kind a clinic sends by text message:

The clinic will call you before your appointment. Bring a list of your medicines. If you cannot come, tell us at least one day early.

Applying the counting rules above gives , , syllables and 105 letters. Two words have three or more syllables — appointment and medicines — so the complex-word count and the polysyllable count are both 2 here. That gives words per sentence and syllables per word.

Substituting into Flesch Reading Ease:

Formula: RE = 206.835 - 1.015 ⁢(8.333) - 84.6 ⁢(1.320) = 86.7

And into Flesch-Kincaid:

Formula: FKGL = 0.39 ⁢(8.333) + 11.8 ⁢(1.320) - 15.59 = 3.2

The full set of six, as the calculator reports them for this passage, is below. Paste the reminder into the box above and you should reproduce every figure exactly.

Look at the spread: 2.5 to 7.8 on a passage of twenty-five plain words. SMOG is the highest because it is the most sensitive to polysyllables and because normalising two polysyllables in three sentences up to a thirty-sentence basis is an aggressive extrapolation on such a short sample. ARI is the lowest because the words here are short in letters even where they are long in syllables. This is exactly why the calculator reports the median rather than any single index, and why it warns you when the sample is too small for the spread to mean anything.

Reading the numbers against a plain-language target

The interpretation table below is Flesch's own, from Table 5 of the 1948 paper. Note the "typical magazine" column: Flesch calibrated the scale against the periodicals of the late 1940s, and the style labels are his. The grade equivalents in the last column come from Flesch's later popular treatment of the same scale rather than from the 1948 regression, so treat them as a convention rather than as a measured result.

For US federal communication there is a statutory backdrop. The Plain Writing Act of 2010 (Public Law 111-274) requires executive agencies to use "writing that is clear, concise, well-organized, and follows other best practices appropriate to the subject or field and intended audience" in documents aimed at the public. The Act does not name a grade level, and the Federal Plain Language Guidelines maintained at plainlanguage.gov are explicit that writing for your actual audience matters more than hitting a formula target — there is no point writing at eighth-grade level for an audience of specialists. Health-literacy practice is where the low targets come from: SMOG at grade 6 or below is the conventional benchmark for patient material, which is why SMOG rather than Flesch-Kincaid is the index most often quoted in that field.

A practical way to read the panel: if the five grade indices cluster within about two grades of each other, the passage is stylistically consistent and the median is a fair summary. If they spread over four or more grades, look at which way. High Fog and SMOG with low ARI means short but jargon-heavy writing — the fix is vocabulary. High ARI and Coleman-Liau with low Fog means long sentences of ordinary words — the fix is punctuation.

Limitations of syllable counting and of the formulas themselves

Syllable counting is a heuristic, not a lookup. This page has no pronunciation dictionary. The rules above handle silent e, consonant plus -le, regular -ed and -es endings, and word-initial y, and they are correct on the great majority of ordinary English words. They are wrong on vowel sequences pronounced as two syllables (poem, creating, diet) unless the word happens to be on the small exception list, wrong on many proper nouns and loanwords, and wrong on technical vocabulary generally. Expect a syllable count within one or two per cent of a dictionary-based count on ordinary prose, and larger errors on specialised text. Any tool that claims exact syllable counts without a dictionary is overstating what it does.

Sentence detection is heuristic too. The abbreviation list is finite. An unlisted abbreviation ending in a full stop will split a sentence in two, which shortens average sentence length and makes the text look easier than it is. Bullet lists without terminal punctuation collapse into one long "sentence", headings run into the paragraph beneath them, and code, URLs and citation strings confuse everything. Strip formatting artefacts before you analyse, or accept that the number is approximate.

The formulas were validated on specific corpora, not on your text. Flesch fitted his equation against the McCall-Crabbs reading lessons and reported a multiple correlation around 0.70 against the 1925 test version. Kincaid and colleagues re-fitted the coefficients against 531 Navy enlisted personnel reading Rate Training Manuals. McLaughlin calibrated SMOG against 100 per cent comprehension rather than the 50 to 75 per cent criterion the Flesch family uses, which is the structural reason SMOG usually returns a grade or two higher than Flesch-Kincaid on the same passage. None of these corpora look like modern web copy, marketing email, legal contracts, or software documentation. The scores remain useful as relative measures within a document set; they are not measurements of a reader's actual comprehension.

What the formulas cannot see. They cannot see whether the passage is organised, whether the reader has the background knowledge to decode a short familiar-looking word like vest in a financial context, whether the pronoun references resolve, or whether the layout supports scanning. They cannot see negation, ambiguity, or an unexplained acronym. A text can be optimised into a high Reading Ease score by chopping every sentence in half and substituting shorter but vaguer words, and it will be worse writing. Treat the score as a prompt to reread, never as a target to hit.

Small samples are unstable. Below roughly 100 words, adding or removing a single sentence moves average sentence length enough to shift the grade level by more than a whole year. SMOG in particular assumes a 30-sentence sample and extrapolates aggressively when given fewer; the calculator flags this rather than hiding it. Grade-level indices are reported with a floor of 0, because a negative school grade has no meaning; the raw equations can and do go negative on very short, very simple text.

Sources used for every constant on this page

Flesch Reading Ease: Rudolf Flesch, "A new readability yardstick", Journal of Applied Psychology 32(3), 1948, pages 221-233 — Formula A on page 225 and the Step 7 restatement on page 229 read R.E. = 206.835 - 0.846 wl - 1.015 sl, with wl in syllables per 100 words; Table 5 on page 230 supplies the interpretation bands reproduced above. This equation and table were read directly from the published article. Flesch-Kincaid Grade Level: J. P. Kincaid, R. P. Fishburne Jr., R. L. Rogers and B. S. Chissom, Derivation of New Readability Formulas (Automated Readability Index, Fog Count and Flesch Reading Ease Formula) for Navy Enlisted Personnel, Naval Technical Training Command, Research Branch Report 8-75, 1975 (ERIC accession ED108134), the primary source for the coefficients 0.39, 11.8 and 15.59. Gunning Fog Index: Robert Gunning, The Technique of Clear Writing, McGraw-Hill, 1952. SMOG: G. Harry McLaughlin, "SMOG grading - a new readability formula", Journal of Reading 12(8), 1969, pages 639-646 (ERIC accession EJ004016); the hand-calculation procedure and the polysyllable definition follow the National Cancer Institute SMOG handout distributed by the Ohio State University Office of Geriatrics and Gerontology. Coleman-Liau: Meri Coleman and T. L. Liau, "A computer readability formula designed for machine scoring", Journal of Applied Psychology 60(2), 1975, pages 283-284. Plain-language guidance: Plain Writing Act of 2010, Public Law 111-274, and the Federal Plain Language Guidelines published by the Plain Language Action and Information Network at plainlanguage.gov/guidelines and digital.gov.

Verification note, stated plainly: the Flesch 1948 equation, its coefficients and Table 5 were checked against the article itself. The Kincaid 1975 report, Gunning 1952 and Coleman and Liau 1975 are cited from the published bibliographic record and from the consistent restatement of their equations across the literature; the original Navy report PDF and the two journal articles are paywalled or unavailable for direct download from this environment, so those three sets of constants have not been read off the primary document here. They are reported as published rather than as independently verified.

Questions writers ask about readability scoring