How the Flesch Reading Ease and Flesch-Kincaid formulas work
Both scores are built from exactly two numbers: the average number of words per sentence, and the average number of syllables per word. Writing them as ASL (average sentence length, in words) and ASW (average syllables per word), the tool computes Flesch Reading Ease as FRE = 206.835 - (1.015 x ASL) - (84.6 x ASW), and the Flesch-Kincaid Grade Level as FKGL = (0.39 x ASL) + (11.8 x ASW) - 15.59. Both results are rounded to one decimal place before they are displayed.
Reading Ease is usually described as a 0 to 100 band where higher is easier, but the formula has no clamp at either end and neither does this tool. A run of six one-syllable words in a single sentence produces 116.1, and a sentence of long abstract nouns can drop well below zero. Treat anything outside 0 to 100 as a signal that the text is unusual rather than as a meaningful score.
Notice that the two formulas read the same two inputs with different weights and opposite signs, so they almost never disagree about direction. If Reading Ease goes up after an edit, the grade level goes down. They are two views of one measurement, not two independent opinions, so there is no value in reporting both as though they corroborate each other.
A worked example: scoring one sentence by hand
Take the sentence “The quick brown fox jumps over the lazy dog.” The tool splits on whitespace and counts 9 words. It counts sentences by looking for one or more of . ! ? followed by a space or the end of the text, which here matches once, so ASL = 9 / 1 = 9.
For syllables it scores each word separately: The 1, quick 1, brown 1, fox 1, jumps 1, over 2, the 1, lazy 2, dog 1. That totals 11 syllables, so ASW = 11 / 9 = 1.222. Substituting gives FRE = 206.835 - (1.015 x 9) - (84.6 x 1.222) = 94.3, and FKGL = (0.39 x 9) + (11.8 x 1.222) - 15.59 = 2.3. The panel reports 94.3, 2.3, 9 words, 1 sentence, 11 syllables, and 1.22 average syllables per word.
Now compare “Our organization leverages synergistic methodologies.” It is shorter at 5 words and 1 sentence, but the tool counts 19 syllables, giving ASW = 3.8 and a Reading Ease of -119.7 with a grade level of 31.2. Five words produced a score far worse than nine words did, which shows how heavily the 84.6 coefficient on syllables dominates the arithmetic. Word length, not sentence length, is what usually moves these numbers.
How this tool counts syllables, and where the heuristic misses
There is no pronunciation dictionary behind the count. The tool lowercases each word, strips everything that is not a letter a to z, returns 1 for any word of three letters or fewer, then removes a final silent-e style ending and counts the remaining runs of one or two vowels, treating y as a vowel. The FAQ puts this vowel-group heuristic at roughly 80 to 90 percent accuracy on ordinary prose, which matches what it does in practice.
The errors run in both directions. It undercounts adjacent vowels that belong to separate syllables: area and idea come back as 2 rather than 3, science and poem as 1 rather than 2, cooperate as 3 rather than 4, and evaluation as 4 rather than 5. It overcounts where a vowel run is silent: business scores 3 instead of 2, and queue and mile score 2 instead of 1. Numerals and symbols are stripped before counting, so 2026 and 15% contribute zero syllables while still counting as words.
On a long piece of ordinary writing these mistakes partly cancel and the averaged result is stable enough to compare drafts. On a short passage they do not cancel. A single misread word in a 20 word sample shifts ASW by 0.05, which moves Reading Ease by about 4 points. If you want a number you can act on, score at least a few hundred words.
How words and sentences are counted, and what breaks the count
Words are whitespace-separated tokens with no further cleaning, so a hyphenated compound like well-known counts as one word, and a bullet character or a stray dash on its own line counts as a word with zero syllables. Pasting a bulleted list or a table therefore dilutes the syllable average slightly.
Sentence detection is the larger hazard, because any period followed by a space ends a sentence. “Dr. Smith went to Washington D.C. yesterday.” is one sentence to a reader and three to the tool, which cuts ASL from 7 to 2.33 and lifts Reading Ease to 71.5. Text heavy in abbreviations, initials, decimal numbers with spaces after them, or ellipses will read as easier than it is.
The opposite failure matters too. If the text contains no terminal punctuation at all, the sentence count falls back to 1, so a 300 word paste with no full stops is measured as one 300 word sentence and scores catastrophically. Headings, list items, and social copy often have no closing punctuation, so add it or strip those lines before scoring. Only the Reading Ease number drives the band label shown beneath it, using the cut-offs 90 and above for very easy, 70 and above for fifth grade, 60 and above for plain English, 30 and above for college, and anything lower for professional.
What a readability score does not measure
These formulas count syllables and sentence length. They do not read. Nothing in the arithmetic inspects vocabulary difficulty, sentence structure, logical order, whether a term was defined before it was used, or whether the text is true. Two sentences with identical word and syllable profiles receive identical scores no matter which one makes sense.
That has a practical consequence in both directions. Jargon-heavy writing can score well: a page of short, familiar-looking technical terms will pass a Reading Ease target while remaining opaque to anyone outside the field, because the formula has no way to know that a short word is a term of art. And genuinely clear writing can score badly, since precise multisyllabic words are often the plainest choice available. Replacing a familiar long word with a vaguer short one improves the score and harms the reader.
The formulas were also calibrated on English, using English syllable structure and mid-century English-language schooling data. The syllable counter here only recognises the letters a to z, so text in German, Korean, Spanish, or any non-Latin script produces a number without meaning rather than a translated equivalent. Use the score as a rough check on one draft against another in English, and use a human reader, or better a reader from the intended audience, to judge whether the writing is actually clear.
Choosing a target score, and editing toward it honestly
A general web audience is usually served by a Reading Ease around 60 to 70, roughly a seventh to ninth grade level. Public health and government plain-language guidance often asks for higher, near 70 to 80. Documentation for practitioners and academic writing routinely sit between 30 and 50, and that is not automatically a fault. Pick the target from who is reading, then hold it steady across a document rather than optimising individual paragraphs.
Because the score responds mostly to word length and secondarily to sentence length, the edits that move it are splitting long sentences at their natural joins, cutting filler that adds syllables without adding meaning, and replacing an abstract noun phrase with a verb. Those edits usually help the reader as well as the number. The edits that only help the number are shortening technical terms into approximations and chopping sentences at points where the connective logic was doing real work.
A useful discipline is to record the score before editing, make the changes you would have made anyway, and check whether the number followed. If the score improved, you have a weak confirmation. If the score got worse but the draft reads better, trust the draft. The analysis runs entirely in your browser and nothing you paste is uploaded, so you can run an unpublished document through it as often as you like.