How We Count Words, Characters and Readability
Every number in the WritePadPro editor and tools comes from one counting engine that runs in your browser. This page documents its rules exactly as they are coded, with a worked example you can reproduce and the limits you should know about.
In short
Words are runs of text separated by spaces (Chinese, Japanese, Thai and similar scripts are split into words). Characters are what you see as one symbol, so an emoji counts once. Sentences end at . ! ? or … followed by a space, with protection for abbreviations, decimals and initials. Reading time uses 238 words per minute, speaking time 150.
One engine for every number
The counting rules live in one JavaScript file that every page loads. The homepage editor, the writing workspace and the counting and analysis tools all call the same functions, so the same text gives the same word, sentence and syllable counts wherever you paste it. The code runs on your device; your text is not sent to a server to be counted (see How Privacy Works).
A few utility tools, such as the Case Converter, show a simpler whitespace word count. It gives the same result except for text written without spaces between words, such as Chinese.
Words
A word is any run of characters between whitespace (spaces, tabs, line breaks, non-breaking spaces and other Unicode spaces). This is the rule most word processors use, and it means punctuation never splits a word.
Chinese, Japanese, Thai, Lao, Khmer and Burmese are written without spaces between words. When a chunk of text contains one of these scripts, the browser’s built-in word segmenter (Intl.Segmenter) splits it into words. The segmenter uses a dictionary that ships with the browser, so results can differ slightly between browsers and versions. In a browser without it, each unspaced chunk counts as one word.
| Text | Words | Why |
|---|---|---|
well-known | 1 | Hyphenated compounds have no space inside. |
don't | 1 | Contractions have no space inside. |
3.14 | 1 | Numbers count as words. |
https://example.com/a-b?x=1 | 1 | A web address without spaces is one word. |
Hello — world | 3 | A dash with spaces on both sides stands alone, so it counts. |
我喜欢写作 | 3 | Segmented as 我 / 喜欢 / 写作 (result from the segmenter in Chrome and Node.js). |
สวัสดีครับ | 2 | Thai is segmented the same way: สวัสดี / ครับ. |
Hello世界 | 2 | Mixed scripts in one chunk are segmented. |
👍🏽 | 1 | An emoji between spaces is a word too. |
Word tokens for keyword and vocabulary lists
Lists of words (keyword density, word frequency, unique words, syllables) need the words themselves, without punctuation. For these, the engine extracts word tokens: runs of letters, accent marks and digits, keeping an apostrophe or hyphen inside a word (“don’t”, “well-known”, “naïve”). Tokens are lowercased and curly apostrophes become straight ones, so “Don’t” and “don't” match. Text in scripts without spaces is split with the same segmenter.
Because tokens contain only letters and digits, their total can differ from the word count: “2.5” is one word but two tokens (“2” and “5”), a web address becomes several tokens, and a lone dash or emoji gives none.
Characters
WritePadPro counts user-perceived characters, which Unicode calls grapheme clusters: what a reader sees as one symbol counts as one. Text with no combining marks, emoji modifiers or joiners is counted by code points; anything more complex goes through the browser’s grapheme segmenter.
- Characters (with spaces) counts everything, including spaces and line breaks. A Windows line break (carriage return + line feed) counts as one character.
- Characters (no spaces) removes every kind of whitespace first: spaces, tabs, line breaks and non-breaking spaces.
| Text | Our count | UTF-16 code units | Note |
|---|---|---|---|
| 👍🏽 (thumbs up, skin tone) | 1 | 4 | Emoji + skin-tone modifier |
| 🇺🇸 (flag) | 1 | 4 | Two regional-indicator letters |
| 👨👩👧 (family) | 1 | 8 | Three emoji joined with zero-width joiners |
| é (e + combining accent) | 1 | 2 | Looks identical to the single character é |
| สวัสดีครับ | 7 | 10 | Thai vowel and tone marks join the letter before them |
Many websites and apps count differently. JavaScript’s length and many web forms count UTF-16 code units, X counts some characters twice, and SMS switches encoding when you use an emoji. Our character limit guides explain each platform’s rule.
Sentences
The text is first split into blocks: paragraphs separated by a blank line, and lines that start a list item (-, *, •, +, 1. or 1)) or a Markdown heading (#). A block always starts a new sentence, so a heading or bullet without a final period is not merged into the next sentence.
Inside a block, a sentence ends at one or more of . ! ? … ‽, optionally followed by closing quotes or brackets, and then a space or the end of the text. The Chinese and Japanese marks 。!? end a sentence even without a following space. These are not treated as sentence ends:
- Periods with no space after them: decimals, web addresses and file names (3.14, example.com).
- Initials: a single capital letter and a period (J. K. Rowling, John F. Kennedy).
- Titles and Latin abbreviations, which never end a sentence: Mr., Mrs., Ms., Mx., Dr., Prof., Rev., Hon., Sgt., Capt., Lt., Col., Gov., Sen., Mt., vs., cf., e.g., i.e., viz., approx. and a few more.
- Dotted abbreviations (U.S., p.m., Ph.D.) unless the next word starts with a capital letter.
- General abbreviations such as etc., Inc., Ltd., Co., Jr., St., Ave., Dept., min. and kg when the next word starts with a lowercase letter or a digit.
- Number abbreviations such as No., Vol., Fig., Ch., pp., Eq. and Ref. when a number follows (“See No. 5”).
- A numbered list marker at the start of a block (“1. First step”).
- A quoted question or exclamation that continues in lowercase (He asked “why?” and left).
- An ellipsis followed by a lowercase word (“Wait… what?” is one sentence).
A fragment with no letters or digits, such as a lone emoji or “:)”, is attached to the sentence before it. Any text that is not empty has at least one sentence.
| Text | Sentences |
|---|---|
| Mr. Smith arrived at 3 p.m. on Friday. He left early. | 2 |
| J. K. Rowling wrote it. Then she rested. | 2 |
| The price rose 3.5% in Q1. Sales fell. | 2 |
| See No. 5 for details. Acme Inc. is next. We use apples, pears, etc. and more. | 3 |
| He asked "why?" and left. | 1 |
| Wait... what? I see... Then he left. | 3 |
| # Heading without period, then a line of text | 2 |
Known limit: the rules look only at the next word. “We met in the U.S. Senate” counts as two sentences because a capital letter follows “U.S.”, and an unknown abbreviation followed by a capitalized name can end a sentence too early.
Paragraphs and lines
- In the tools, paragraphs are separated by a blank line (two or more line breaks with only whitespace between them). The Paragraph Counter can also count every line break as a new paragraph.
- In the editor, each non-empty line counts as a paragraph, because every paragraph you create with Enter is its own line.
- Lines are counted the way a plain-text editor numbers them: line breaks + 1. The Word Counter’s “Lines” figure shows only non-empty lines.
Syllables
English spelling does not map exactly to pronunciation, so syllables are estimated with rules plus a dictionary of common exceptions. For each word token:
- Hyphenated compounds are split and their parts added up (state-of-the-art = 4).
- Accents are removed (café → cafe), but a final é or è keeps its own syllable.
- A list of 110 common irregular words is checked first (for example business = 2, every = 2, idea = 3, people = 2).
- Words of one or two letters count as one syllable.
- Otherwise, groups of vowels are counted. “Y” counts as a vowel except at the start of a word before a vowel (yes) and between two vowels (player).
- One is subtracted for a silent final “e” (cake), unless it forms “-le” or “-re” after a consonant (table, centre); for “-es” and “-ed” endings that add no sound (makes, jumped, but not boxes, judges, wanted, handled); for a silent “e” before -ly, -ful, -less, -ness or -ment (lately); and for “-que” and “-gue” (unique, league).
- One is added where two vowels are pronounced separately: “-ing” after a vowel (flying), “-ier/-iest” (happier), “ia” and “iu” (media), and endings such as -ious, -eous, -uous, -iate, -ual, -uate, -uity, -uent, -iety, -ient, -ism, -io and -eo (various, video), plus words starting with “bio”.
Every word with at least one letter has at least one syllable. Tokens without letters, such as “2024”, have none. Other scripts get rough counts so the formulas still work: vowels in Cyrillic, vowel groups in Greek, one per character in Chinese, Japanese and Korean, and one per word elsewhere.
Known limits: the rules miss some words. “Aisle” gets 2 (dictionaries say 1), “colonel” 3 (2) and “Seattle” 2 (3). Words like “fire” and “hour” vary by accent; names and borrowed words are the hardest. Check individual words with the Syllable Counter.
Reading and speaking time
Reading time = word count ÷ 238 × 60 seconds, rounded to the nearest second. 238 words per minute is the average silent reading rate for English non-fiction reported by Marc Brysbaert in “How many words do we read per minute? A review and meta-analysis of reading rate” (Journal of Memory and Language, 2019). Real readers vary a lot with the text and their purpose, so treat it as an average, not a prediction.
Speaking time = word count ÷ 150 × 60 seconds. 150 words per minute is our standard for a comfortable presentation pace. Speakers differ, so the Speaking Time Calculator lets you set your own pace. The editor also shows a presentation time at 130 words per minute and a handwriting time at 13 words per minute; both are rough estimates.
Page estimates
Our baseline is Times New Roman 12 pt on US Letter paper with 1-inch margins: 500 words per single-spaced page and 250 per double-spaced page. Other settings scale that baseline:
words per page = round(500 × font factor × page-area factor ÷ line spacing)
Line spacing divides by 1, 1.15, 1.5 or 2. The page-area factor is the text area divided by that of US Letter with 1-inch margins (A4 with 1-inch margins = 1.04). Font factors compare each font’s average character width and line height with Times New Roman 12 pt. Pages are shown to one decimal place.
| Font | Factor | Single | 1.5 | Double |
|---|---|---|---|---|
| Times New Roman 12 pt | 1.00 | 500 | 333 | 250 |
| Arial 11 pt | 1.08 | 540 | 360 | 270 |
| Arial 12 pt | 0.91 | 455 | 303 | 228 |
| Calibri 11 pt | 1.10 | 550 | 367 | 275 |
| Calibri 12 pt | 0.92 | 460 | 307 | 230 |
| Georgia 12 pt | 0.92 | 460 | 307 | 230 |
| Verdana 10 pt | 1.07 | 535 | 357 | 268 |
| Verdana 12 pt | 0.74 | 370 | 247 | 185 |
| Courier New 12 pt | 0.68 | 340 | 227 | 170 |
| Garamond 12 pt | 1.06 | 530 | 353 | 265 |
Handwritten pages use 200, 250 or 300 words per side for large, average and small handwriting. Real page counts also depend on headings, paragraph spacing and the program you use, so these are estimates. Try other settings in the Words to Pages Calculator or read the writing guides for common word counts.
Readability formulas
The six scores use the word count, the sentence count and the syllables of the word tokens described above. Results are rounded to two decimals (the Readability Checker shows one). The formulas were designed for English prose; scores for other languages, lists, code or a sentence or two are not meaningful.
| Score | Formula | How to read it |
|---|---|---|
| Flesch Reading Ease (Flesch, 1948) | 206.835 − 1.015 × (words ÷ sentences) − 84.6 × (syllables ÷ words), limited to 0–100 | Higher is easier. 60–69 is labeled “Standard”. |
| Flesch–Kincaid Grade (Kincaid et al., 1975) | 0.39 × (words ÷ sentences) + 11.8 × (syllables ÷ words) − 15.59, minimum 0 | US school grade: 8 means an eighth-grade reader. |
| Gunning Fog (Gunning, 1952) | 0.4 × ((words ÷ sentences) + 100 × (complex words ÷ words)) | Years of schooling. Complex = 3+ syllables. We count every such word; Gunning’s guidance also excluded names, compounds and some suffixes. |
| Coleman–Liau (Coleman & Liau, 1975) | 0.0588 × L − 0.296 × S − 15.8; L = letters per 100 words, S = sentences per 100 words | US grade. Uses letters, not syllables; letters in any alphabet count. |
| SMOG (McLaughlin, 1969) | 3 + √(words of 3+ syllables × 30 ÷ sentences); needs at least 3 sentences | Years of schooling. Tools using the refined formula (1.043 × √… + 3.1291) show slightly different values. |
| Automated Readability Index (Senter & Smith, 1967) | 4.71 × (characters ÷ words) + 0.5 × (words ÷ sentences) − 21.43; characters = letters and digits | US grade. Uses characters, not syllables. |
| Score | Difficulty | Reading level |
|---|---|---|
| 90 and above | Very Easy | 5th grade |
| 80–89 | Easy | 6th grade |
| 70–79 | Fairly Easy | 7th grade |
| 60–69 | Standard | 8th–9th grade |
| 50–59 | Fairly Difficult | 10th–12th grade |
| 30–49 | Difficult | College |
| Below 30 | Very Difficult | Graduate |
For practical advice on raising a score, read How to Improve Your Readability Score or paste your text into the Readability Checker.
Keywords, frequency and vocabulary
- Keyword density = occurrences of a word ÷ total word tokens × 100. Keyword lists leave out one-letter words and 161 common English stop words (the, and, of, is…). The Keyword Density Checker also lists two- and three-word phrases that appear at least twice; phrases never cross sentence punctuation, commas, brackets, quotes, dashes or line breaks. Its colors (1–2% green, above 2% amber, above 3% red) are a rough editing guide, not a search-engine rule.
- Word frequency uses the same tokens and the same formula, with stop words included unless you hide them.
- Unique words = the number of different tokens, ignoring capitalization.
- Vocabulary richness (type-token ratio) = unique words ÷ total tokens × 100. The Vocabulary Analyzer adds root TTR (unique ÷ √tokens), corrected TTR (unique ÷ √(2 × tokens)), words used only once, and lexical density (tokens that are not stop words ÷ all tokens). The type-token ratio falls as a text gets longer, so compare texts of similar length.
- Average word length = characters in all tokens ÷ number of tokens. Average sentence length = words ÷ sentences.
Editor highlights
The editor marks style issues in place. Each sentence gets a grade from the Automated Readability Index (rounded to a whole number), and the rules below are applied to English text. For text that does not look like English (judged by the share of common English function words), only sentence length is checked: 20+ words is hard and 30+ is very hard.
| Highlight | Rule |
|---|---|
| Hard to read | A sentence of 14 or more words at grade 10–13. |
| Very hard to read | A sentence of 14 or more words at grade 14 or higher, or any sentence of 30 or more words. |
| Adverb | A word of five or more letters ending in “-ly”, except 109 common words that are not adverbs (family, only, early, likely, reply, supply…). |
| Passive voice | A form of “be” (am, is, are, was, were, be, been, being), optionally followed by one adverb or a word such as “not” or “often”, then a past participle (a word ending in “-ed” or one of 98 irregular forms such as written, given, built). Participles that usually act as adjectives (tired, excited, interested and 62 more) count only when “by” follows within three words. |
| Simpler alternative | One of 99 words (utilize → use) or 26 wordy phrases (in order to → to) with a shorter alternative. |
| Cliché | One of 180 stock phrases; the longest match wins. |
| Qualifier | One of 41 words or phrases that weaken a statement (very, really, quite, sort of, I think…). |
| Repetition | The same word twice in a row (“the the”; “had had” and “that that” are allowed), or a content word of three or more letters used three or more times within 40 words. |
| Long paragraph | A paragraph of more than 150 words. |
Highlights are prompts, not errors: a passive sentence or an adverb is often the right choice. In documents longer than 60,000 words, live highlights pause (an “Analyze now” button runs them) while the statistics stay live.
The editor’s rhythm score (0–100) measures sentence-length variety once you have three or more sentences. It combines the spread of sentence lengths (70%, from the coefficient of variation: 0.5–0.9 scores best) with how often neighboring sentences are within two words of each other (30%). Scores of 80+ read as “Lively rhythm”, 60+ “Good variety”, 40+ “A little uniform” and below that “Monotonous”.
Grammar and spelling
Grammar and spelling checks use Harper (version 2.10.0), an open-source checker by Automattic. It is a WebAssembly program of about 16 MB that your browser downloads once from our server and then runs in a background worker on your device. It supports American, British, Canadian and Australian English and checks your text paragraph by paragraph. Words you add to your personal dictionary are stored in your browser.
The Grammar & Spell Checker adds its own rule layer: a list of more than 250 common misspellings plus patterns for grammar, capitalization, punctuation and spacing. Rule-based checkers catch common mistakes but not every error, and they can flag correct names or technical terms.
AI tokens
The AI Token Counter runs OpenAI’s o200k_base encoding (from the open-source gpt-tokenizer library, version 4.0.0) in your browser, so counts for the OpenAI models that use this encoding are exact. For other providers the tool shows a range estimated from characters: 3 to 4 characters per token for Claude, 3.5 to 4.5 for Gemini and Llama, 3.2 to 4 for Mistral, and 0.8 to 1.5 tokens per Chinese, Japanese or Korean character. Treat those as estimates.
AI-pattern check
The AI Content Detector does not decide who wrote a text. It measures eight writing patterns that are common in text from large language models, scores each from 0 to 100, and combines them in a weighted average. It needs at least 80 words.
| Signal | What is measured | Weight |
|---|---|---|
| Overused phrases | Weighted matches from a list of about 50 phrases (“a testament to”, “multifaceted”, “embark on a journey”…) per 1,000 words. English only. | 2 |
| Burstiness | Variation in sentence length (coefficient of variation 0.25 or less scores 100, 0.70 or more scores 0). Needs 5+ sentences. | 1.5 |
| Transitions | Share of sentences that open with a transition such as “However” or “Additionally” (8% or less = 0, 35% or more = 100). English only. | 1 |
| Sentence openers | Share of the most common first word (20% or less = 0, 45% or more = 100). Needs 6+ sentences. | 0.75 |
| Hedging | Hedges per 100 words (1.5 or less = 0, 4.5 or more = 100). English only. | 0.75 |
| Punctuation | Em dashes plus half the semicolons, per 1,000 words (3 or less = 0, 15 or more = 100). | 0.75 |
| Paragraph uniformity | Variation in paragraph length (0.15 or less = 100, 0.60 or more = 0). Needs 3+ paragraphs. | 0.75 |
| Vocabulary diversity | Average share of different words in each 50-word window (0.66 or less = 100, 0.80 or more = 0). | 0.5 |
What it can’t do: no detector can prove that a text was written by AI. Formal, technical, templated or non-native English writing can score high, and edited AI text can score low. Short texts give unstable results, and three signals are skipped for languages other than English. Never use the score as the only evidence against anyone.
Worked example
We ran this paragraph through the same code your browser runs:
Dr. Rivera's team spent 2.5 weeks testing the new checkout page. It wasn't easy! The well-known layout failed twice, so they rewrote the copy and cut every long sentence. Customers noticed the change immediately. Was it worth the extra week? The data said yes.
| Measure | Result | How it was reached |
|---|---|---|
| Words | 44 | “2.5”, “well-known”, “wasn't” and “Rivera's” are one word each. |
| Characters (with spaces) | 260 | Every letter, digit, space and punctuation mark. |
| Characters (no spaces) | 217 | 260 minus 43 spaces. |
| Sentences | 6 | “Dr.” is a title and “2.5” has no space after the period. |
| Word tokens | 45 | “2.5” becomes the tokens “2” and “5”. |
| Syllables | 63 | For example ri-ve-ra’s 3, im-me-di-ate-ly 5; the digits have none. |
| Words of 3+ syllables | 3 | Rivera’s, customers, immediately. |
| Reading time | 11 sec | 44 ÷ 238 × 60 = 11.09. |
| Speaking time | 18 sec | 44 ÷ 150 × 60 = 17.6. |
| Flesch Reading Ease | 78.26 | 206.835 − 1.015 × (44 ÷ 6) − 84.6 × (63 ÷ 44): Fairly Easy, 7th grade. |
| Flesch–Kincaid Grade | 4.17 | 0.39 × 7.33 + 11.8 × 1.43 − 15.59. |
| Gunning Fog | 5.66 | 0.4 × (7.33 + 100 × 3 ÷ 44). |
| Coleman–Liau | 7.29 | 203 letters and 6 sentences per 44 words. |
| SMOG | 6.87 | 3 + √(3 × 30 ÷ 6). |
| Automated Readability Index | 4.18 | 205 letters and digits per 44 words. |
| Unique words | 39 | “the” appears six times and “it” twice. |
| Vocabulary richness | 86.67% | 39 ÷ 45 × 100. |
| Keyword density | 2.22% | Each content word appears once: 1 ÷ 45 × 100. |
| Editor highlights | 1 | Adverb “immediately”. The longest sentence (15 words) is at grade 9, below the “hard” threshold. |
| OpenAI tokens (o200k) | 58 | Counted with the tokenizer the AI Token Counter uses. |
| Pages (double-spaced) | 0.2 | 44 ÷ 250 = 0.176, shown to one decimal. |
Paste the paragraph into the Word Counter or the Readability Checker to check these numbers yourself.
Why another counter may show a different number
Two honest counters can disagree because they answer slightly different questions. The usual reasons:
- What counts as a word. Tools differ on hyphenated words, numbers with decimals, web addresses, “and/or”, dashes between spaces and emoji.
- Languages without spaces. Some tools count each Chinese or Japanese character as a word, some count a whole unspaced run as one word, and others use a dictionary, as we do.
- What counts as a character. Visible characters, Unicode code points, UTF-16 code units and bytes all give different totals for emoji, accented letters and many scripts. Some tools count a Windows line break as two characters.
- Hidden characters. Text copied from the web or a PDF can contain non-breaking spaces, zero-width spaces or soft hyphens. The Invisible Character Detector shows them.
- What is included. Word processors can include or leave out footnotes, text boxes, headers and captions. We count only the text you paste or type.
- Sentence rules. Abbreviations, ellipses, quotes and list items are handled differently from tool to tool, which changes sentence counts and every readability score built on them.
- Syllables and formula versions. Dictionary lookups and rule-based estimates disagree on some words, and SMOG and Gunning Fog exist in more than one version.
- Assumed speeds and rounding. Reading-time tools use different reading speeds, and some round up to whole minutes or whole pages where we show seconds and one decimal.
- Copy-and-paste side effects. Text copied from PDFs often gains line breaks inside sentences or hyphens at line ends, which change word and sentence counts.
If you think one of our numbers is wrong, send us the text and the result you expected. We check it against the rules on this page and fix the engine or this page if either is wrong (see our editorial policy); changes to how we count are listed in the changelog.
Frequently asked questions
Why does WritePadPro count “well-known” as one word?
Words are separated by spaces, so a hyphenated compound, a contraction such as “don’t”, a number such as 3.14 and a web address each count as one word. A dash with a space on both sides (“yes — no”) counts as a word of its own.
Does an emoji count as one character?
Yes. WritePadPro counts the characters you see (grapheme clusters), so 👍🏽, 🇺🇸 and a family emoji each count as one character, even though they take 4 to 8 code units in memory. Social networks and SMS can count them differently; see Character Limits.
Why do readability scores differ between tools?
Every readability formula depends on syllable and sentence counts, and tools count both differently. Some also use other versions of a formula: WritePadPro uses the short SMOG formula (3 + √…) and treats every word of three or more syllables as complex in the Gunning Fog index.
What reading speed does WritePadPro use?
238 words per minute, the average silent reading rate for English non-fiction in Marc Brysbaert’s 2019 meta-analysis of reading rate. Speaking time uses 150 words per minute. The Reading Time Calculator also shows times at 200 and 300 words per minute.