How we count

Most counters give you a number and never say what they measured, which is why three tools can report three different lengths for the same paragraph. This page is the full specification for ours. It is the single source of truth for every count on the site, and if the tool ever disagrees with what is written here, the tool has a bug.

Characters

A character is a grapheme cluster — the unit a person would point at and call one character.

• An emoji counts as 1, however many code points it is built from. A family emoji made of four people joined by zero-width joiners is 1, not 7.
• An accented letter counts as 1 whether it was typed as a single precomposed character or as a base letter plus a combining mark. Text is normalised before counting.
• A line break is 1 character. Windows-style CRLF endings are normalised so they are not counted as 2.
• A tab is 1 character.

Characters excluding spaces removes every space, tab and line break, and nothing else. Punctuation stays in.

Words

A word is a run of letters or digits, held together across apostrophes, hyphens and internal periods.

• don't is 1 word, with either a straight or a curly apostrophe.
• state-of-the-art is 1 word. Splitting on hyphens would make it 4.
• 3.14 and 1,000,000 are 1 word each.
• U.S.A. is 1 word.
• A URL or an email address is 1 word.
• An em dash or en dash used as punctuation is not a word, and does not join the words either side of it.
• A standalone symbol — &, •, § — is not a word.
• A standalone digit is a word. Chapter 7 is 2 words.

Sentences

A sentence ends at . ! ? or an ellipsis when followed by whitespace or the end of the text, and at any line break.

• Abbreviations do not end sentences. The list includes Dr., Mr., Mrs., Ms., Prof., Sr., Jr., St., Mt., Ave., Rd., Inc., Ltd., Co., Fig., No., vs., approx., e.g., i.e., etc., cf., et al., a.m., p.m.
• Initials are held together. J. R. R. Tolkien is inside one sentence.
• Decimals and version numbers are left alone. 3.14, v2.1.
• A semicolon does not end a sentence. It joins independent clauses, which is its function.
• A line break does end one. Headings, bullet points and lines of verse are counted as sentences even without terminal punctuation.
• A closing quotation mark or bracket after terminal punctuation does not prevent the sentence from ending.

The line-break rule is our largest single divergence from word processors, and it is deliberate: without it, a bulleted document reports one enormous sentence and a meaningless words-per-sentence average.

Paragraphs

A paragraph is a block of text separated from its neighbours by one or more blank lines. A single line break inside a block does not start a new paragraph. Empty blocks are not counted.

Letters, vowels and consonants

• Letters are Unicode letter characters. Digits, punctuation, spaces and emoji are not letters.
• Vowels are A, E, I, O, U in either case, plus their accented forms (à é î ö ũ and the rest). Y is assigned by the rule you choose.
• Consonants are every letter that is not a vowel.

Vowels plus consonants equals letters, exactly. It does not equal the character count, because spaces, digits and punctuation are letters in nobody's definition.

The three Y rules

• Decide per word (default). Y is a consonant when a vowel follows it — yes, beyond, canyon — and a vowel everywhere else: rhythm, happy, my.
• Always vowel. Every Y is counted as a vowel. Common in phonics teaching for young children, where the simpler rule is the point.
• Always consonant. Every Y is counted as a consonant. The traditional English alphabet position, and what most simple counters do implicitly.

Is Y a vowel? goes through the cases where all three rules are defensible.

Syllables

Syllables are estimated, not counted. The algorithm counts vowel groups, then applies corrections for silent terminal e, for -le endings, for common suffixes, and for vowel pairs that form a single nucleus. It is right on the large majority of ordinary English words and wrong on some loanwords, names and unusual spellings.

No syllable counter that does not consult a pronunciation dictionary is exact, and one that does still has to choose an accent. Treat the figure as a good approximation and see how to count syllables for where it fails.

Reading time and speaking time

• Reading time assumes 238 words per minute, the mean for adults reading English non-fiction for comprehension, from Brysbaert's 2019 meta-analysis of 190 studies.
• Speaking time assumes 150 words per minute, a comfortable presentation pace.

Both are averages across readers and texts, and the spread around them is wide. See reading time and speaking time.

Readability scores

Flesch Reading Ease and the Flesch–Kincaid Grade Level are computed from our own word, sentence and syllable counts using the standard formulas. Because the syllable count is an estimate and our sentence rules differ from other tools', our scores will differ from theirs — usually by less than a grade level, occasionally by more on heavily bulleted text. Readability scores explained covers what the numbers mean and how far to trust them.

Other figures

• Unique words are counted case-insensitively, so The and the are one word.
• Longest word is measured in characters, using the word rules above, so a hyphenated compound can win.
• Average word length is characters per word, excluding spaces.
• Most repeated words excludes a small stop list of function words, because a list reporting that the is your most frequent word is not information.
• Letter frequency is case-insensitive and covers A–Z only; accented letters are folded to their base letter for the chart.

Changes to these rules

If a rule on this page changes, the change is noted here with a date, because a count that silently changes its mind is worse than one that was wrong consistently. Nothing has changed since the site launched.