Free Word & Character Counter Online

Paste or type your text below to instantly see word count, character count, sentences, paragraphs, and estimated reading time.

Words

Characters

Chars (no spaces)

Sentences

Paragraphs

Reading Time

Speaking Time

Unique Words

Readability

About This Tool

This free word counter is perfect for students, writers, bloggers, and social media managers. Instantly count words for essays, tweets, Instagram captions, and more. Reading time is based on an average reading speed of 200 words per minute.

Common Word Count Limits

Knowing common limits helps you hit your target:

Twitter/X post · 280 characters
Instagram caption · 2,200 characters
Meta description (SEO) · 155-160 characters
College essay · typically 250-650 words
Blog post (SEO) · 1,500-2,500 words for ranking
Short story · 1,000-7,500 words

Frequently Asked Questions

How is reading time calculated?

We use the standard average of 200 words per minute. This is a widely accepted average for adult reading speed of non-technical content.

What counts as a word?

Any sequence of characters separated by whitespace counts as one word. Hyphenated words like "well-known" count as one word. Numbers and abbreviations also count.

Is my text saved anywhere?

No. Everything happens in your browser. Your text is never sent to any server or stored anywhere. Close the tab and it is gone.

What counts as a word, exactly?

Microsoft Word, Google Docs, Pages, the Unix wc command and almost every browser-based counter agree on a single rule: a word is a maximal run of non-whitespace characters separated from neighbouring runs by whitespace. The POSIX standard for wc defines a word as a non-zero-length sequence of printable characters delimited by white space, the same rule a four-line JavaScript implementation (text.trim().split(/\s+/).length) produces. It's purely mechanical: it doesn't understand morphology, punctuation, or language. "Hello world" is two words; "Hello,world" with no space is one. "Don't" is one word; "do n't" is two.

A linguist would push back. The Oxford English Dictionary distinguishes between the graphemic word (a sequence of letters bounded by spaces) and the lexeme (an abstract dictionary entry). The two diverge in common cases:

Hyphenated compounds. "State-of-the-art" is one word, there's no whitespace inside the token. Microsoft Word, Google Docs and this counter all agree.
Contractions. "Don't," "won't," "it's," "you'd've" all count as one word each. The apostrophe is treated as part of the token, matching standard English usage.
Numbers and decimals. "3.14" is one; "3 . 14" is three. "1,000,000" is one; "1 000 000" (European convention) is three.
Acronyms. "U.S.A." is one word; "U S A" is three. "NATO" is one. "Ph.D." is one.
URLs and emails. "https://absolutool.com/tools" is one word in any whitespace-split counter.

For ordinary prose the disagreement between this rule and a careful human counter is usually under one percent. For text that is highly compound, hyphenated or numerical it can reach 5–15%. Beyond the whitespace rule, Unicode publishes a formal Word Boundary algorithm (UAX #29), a 14-rule state machine that recognises word boundaries based on character categories. JavaScript exposes it via Intl.Segmenter with {granularity: 'word'}, and that's the only correct path for counting words in Chinese, Japanese, Thai or Khmer without a dictionary.

Languages that don't use spaces

Chinese, Japanese (in pure kanji or kanji + hiragana), Thai, Lao, Khmer, Burmese and Tibetan write without inter-word spaces, the convention is called scriptio continua. A whitespace-tokeniser run on a Chinese paragraph returns 1, the entire paragraph as a single "word." For these scripts the meaningful metric is the character count, not the word count.

In professional translation, the rule of thumb is roughly one Chinese character ≈ 1.5–2.0 English words; two Japanese characters ≈ one English word; one Korean character ≈ 0.55 English words. Chinese-to-English translators typically bill by the source character count for the Chinese half and the target word count for the English half. Korean, uniquely among East Asian scripts, does use word-spacing (띄어쓰기, ttuieosseugi), so whitespace tokenisation works, but the spacing rules are notoriously difficult and even native speakers disagree on edge cases.

A short history of paying writers by the word

Per-word payment originated with the mass-circulation magazines of the late nineteenth century. By the 1920s and 1930s the pulp magazines, Black Mask, Weird Tales, Astounding, Argosy, paid by the word at rates from a quarter-cent to two cents. Raymond Chandler famously got a penny per word from Black Mask in the 1930s. The model survives today: penny-stock content mills pay $0.01–$0.05 per word, mid-tier trade magazines pay $0.30–$0.75, and top-tier US glossies (The New Yorker, The Atlantic, Wired, Harper's) pay $1–$3 per word, occasionally higher for high-profile features.

Because the cap is contractual, knowing the exact word count is income. A 1,500-word feature at $1 per word is $1,500; at 1,400 it's $1,400. Editors enforce caps, a 2,000-word commission means 1,950–2,050, not 2,400, and missing the cap can mean a flat-rate cut or a kill-fee.

Novel word counts and the daily-word habit

Publishing has remarkably consistent target word counts for adult novels. Adult literary fiction lands at 80,000–100,000 words; commercial thriller, mystery and romance at 70,000–90,000; epic fantasy and sci-fi at 100,000–120,000 (occasionally 150,000+). Young Adult is 50,000–80,000, Middle Grade 25,000–55,000, picture books 100–500, novellas 17,500–40,000, short stories 1,000–7,500, flash fiction under 1,000. Agents will reject queries for first novels far outside these ranges regardless of quality, the conventional ranges encode reader expectation and printing-press economics.

National Novel Writing Month, founded in 1999 by Chris Baty, sets a 50,000-word target for November, roughly the length of The Great Gatsby (50,061) and Slaughterhouse-Five (49,459). That works out to 1,667 words a day for thirty days. Stephen King recommends 2,000 words a day in On Writing; Anthony Trollope wrote 3,000 a day before going to his post-office job; Graham Greene reportedly stopped at exactly 500 words a day; Mark Twain aimed for 1,400. A daily-word habit is what turns "I have an idea" into "I'll finish in N days." A 90,000-word novel at 1,000 words a day is 90 days.

Academic and journal word counts

For US and UK higher education the typical assignment lengths run: high-school essay 300–1,000 words; university essay (humanities) 1,500–3,000; term paper 3,000–6,000; honours thesis 12,000–20,000; master's thesis 20,000–40,000; PhD dissertation 80,000–100,000 in the humanities (often shorter, 40,000–60,000, in the sciences). The Common App personal statement is capped at 650 words, the application portal enforces it. The UCAS personal statement in the UK is capped at 4,000 characters or 47 lines, whichever is shorter. Most universities enforce essay limits with a ±10% tolerance.

Journal articles run from a 500–1,500-word letter, a 1,500–3,000-word brief, a 4,000–8,000-word standard research article, to a 6,000–12,000-word review. Nature Letters cap at 3,000 words; JAMA Original Investigation at 3,000 with a three-paragraph abstract; PNAS Direct Submission at six printed pages, roughly 6,000 words.

Social-media character limits worth memorising

X/Twitter post: 280 characters (originally 140 at launch in March 2006; doubled to 280 in November 2017). X Premium subscribers can post up to 25,000 characters since February 2023. URLs always count as 23 characters regardless of underlying length, emojis count as 2, CJK characters count as 2.
Bluesky: 300 characters.
Threads (Meta): 500 characters.
LinkedIn post: 3,000 characters; the first ~210 (desktop) or ~140 (mobile) show before "see more."
Instagram caption: 2,200 characters with a 125-character preview.
TikTok caption: 4,000 characters.
SMS GSM-7: 160 characters per segment; UCS-2 (for non-Latin alphabets) drops to 70.
Google meta description: roughly 155–160 characters before truncation on desktop, ~120 on mobile.
Google title-tag: roughly 50–60 characters or about 600 pixels on desktop.

Sentence segmentation is harder than splitting on a period

The naive rule, split on ., !, ?, fails on every realistic English text within the first paragraph. "Dr. Smith met Mr. Jones at U.S. Steel" is one sentence; the naive split returns four. "J. R. R. Tolkien wrote The Hobbit" is one; naive returns four. "The pH of 7.4 is normal" is one; naive returns two. Domain names ("Visit absolutool.com.") and emails create false positives; ellipses ("Wait...what?") create false negatives.

Robust segmentation needs either a curated abbreviation list (the approach used by NLTK's Punkt segmenter, which Kiss and Strunk's 2006 paper trained on a million-word corpus to F1 ≈ 0.97) or a statistical model. For a browser-based counter, the practical compromise is splitting on [.!?]+ followed by whitespace and a capital letter, with a small abbreviation blacklist (Mr|Mrs|Dr|Prof|Sr|Jr|St|Inc|Ltd|U.S|e.g|i.e|etc|vs). Even with that, accuracy on noisy text caps around 90%. Take the sentence count as approximate.

Readability formulas in one paragraph each

Flesch Reading Ease (1948). Rudolf Flesch's formula: 206.835 − 1.015 × (words/sentences) − 84.6 × (syllables/words). Higher means easier. 0–30 is "very difficult, college graduate"; 60–70 is "plain English, eighth-to-ninth grade"; 90–100 is "very easy, fifth grade." Reader's Digest targets ~65; Time targets ~52; the Harvard Law Review runs around 30.

Flesch-Kincaid Grade Level (1975). Commissioned by the US Navy from J. Peter Kincaid for technical-manual readability: 0.39 × (words/sentences) + 11.8 × (syllables/words) − 15.59. Output is a US grade level (8.0 = eighth grade). Now embedded in Microsoft Word's spell-check and the most-cited readability metric in copywriting.

SMOG (1969). Harry McLaughlin's "Simple Measure of Gobbledygook" counts only words of three or more syllables in ten consecutive sentences from start, middle and end of the text. Preferred by the US National Institutes of Health for healthcare patient leaflets because it correlates better than Flesch-Kincaid with actual comprehension on health text.

Gunning Fog (1952). Robert Gunning's index: 0.4 × ((words/sentences) + 100 × (complex_words/words)). The Wall Street Journal targets Fog ~11; the Times of London ~14. Gunning's claim was that any business writing above Fog 12 would lose readers.

Reading and speaking time, where the numbers come from

This counter shows reading time at 200 words per minute and speaking time at 130 words per minute. The 200-wpm figure is conservative relative to Marc Brysbaert's 2019 meta-analysis of 190 studies, which converged on 238 wpm for non-fiction and 260 wpm for fiction (Journal of Memory and Language). Erring slightly on the slow side is the user-friendly choice, better that a "five minute read" be a four-minute read than a seven-minute read. The 130-wpm speaking rate matches conversational speech; formal speeches average 100–130 wpm. A twenty-minute keynote is roughly 2,600 spoken words. Medium's "X min read" feature, introduced in 2014 by Mike Sall, was originally calibrated at 275 wpm.

Why your count differs from Word's

Three common causes: (1) Microsoft Word and Google Docs treat hyphenated words as one (matching this counter), but some PHP-based counters split on every non-letter and return two for "well-known." (2) Pasting from Word brings invisible characters, curly quotes, non-breaking spaces (U+00A0), soft hyphens (U+00AD), that can fragment or fuse tokens depending on the tool. (3) Unique-word counts are case-sensitive by default, so "The" and "the" count as two separate unique words; stripping case and stemming ("running" → "run") would lower the unique count by roughly 15% on typical prose.

If your Chinese or Japanese text shows zero (or one) words, that's the whitespace-tokeniser doing what it's told, switch your attention to the character count, which is the meaningful metric for those scripts. If "3,000" counts as one word it's because the comma is non-whitespace and the token is contiguous; that's correct behaviour and matches Word.