Word Counter
Words, characters with and without spaces, sentences, paragraphs and unique words — plus the words you lean on most. The rules used to decide what counts as a word are published below, because that is the only reason two counters ever disagree.
Paste your text
Up to 200,000 characters. Counting happens on this device as you type — but note that the share link carries your text in the URL, so do not share a link to a confidential draft.
Ranked with the common words in, the top of any English list is always “the, of, and, to”. Skipping them is what makes the ranking about the subject matter. It changes only the table — the word count itself always counts every word.
330 characters with spaces, 271 without. About 7 sentences across 1 paragraph, and 46 of the words are distinct.
- Characters (with spaces)
- 330
- Characters (without spaces)
- 271
- Sentences (approximate)
- 7
- Paragraphs
- 1
- Unique words
- 46
- Average word length
- 4.21 characters
- Average sentence length
- 8.86 words
| # | Word | Times | Share of all words |
|---|---|---|---|
| 1 | one | 3 | 4.8% |
| 2 | word | 3 | 4.8% |
| 3 | counting | 2 | 3.2% |
| 4 | like | 2 | 3.2% |
| 5 | two | 2 | 3.2% |
| 6 | words | 2 | 3.2% |
| 7 | 1,234 | 1 | 1.6% |
| 8 | alone | 1 | 1.6% |
| 9 | between | 1 | 1.6% |
| 10 | compound | 1 | 1.6% |
What counts as a word, in four decisions
Counting words sounds like counting and is really a series of judgement calls. Is a hyphenated compound one word or two? Is a contraction? What happens to the two words either side of an em dash with no space around it? Every counter answers these, most of them silently, and that is why the same paragraph pasted into three tools comes back three slightly different sizes. Here are the answers this one uses.
joiners → spaces • split on whitespace • trim the edges • keep what has a letter or a digit- Joiners
- Em dashes, en dashes, double hyphens and ellipses become spaces first — they belong to neither word they sit between
- Split
- Whitespace separates, and nothing else does. A single hyphen is not a separator, which is what keeps “well-known” whole
- Trim
- Leading and trailing characters that are neither letters nor digits come off, so “(hello),” is the word hello
- Keep
- What still has a letter or a digit in it is one word. A lone dash or a row of dots is none
Contractions, hyphenated compounds and numbers therefore each come to one word, because the character that might have split them — an apostrophe, a hyphen, a comma, a decimal point — is interior rather than at an edge. Sentences are counted by looking for runs of full stops, question marks and exclamation marks, and a paragraph is any line with something on it. Characters are counted as UTF-16 code units, so anything outside the Basic Multilingual Plane, most emoji included, counts as two.
One sentence, counted by hand
Take: The well-known author didn’t stop—she wrote 1,234 words.
The em dash becomes a space, so “stop—she” separates into two. Splitting on whitespace gives eleven tokens; trimming the edges turns “words.” into “words” and leaves nine that still contain a letter or a digit: The, well-known, author, didn’t, stop, she, wrote, 1,234, words. Nine words.
The rest follows from the same pass. Fifty-six characters with spaces, of which seven are spaces, so forty-nine without. One sentence, because there is one full stop. Forty-seven characters sit inside the nine words, so the average word is 5.22 characters long — a figure that counts the hyphen in “well-known” and the apostrophe in “didn’t”, because they are part of the word as it was written. A counter that split the compound would report twelve words at 4.25 characters each. Same sentence, different rules, different answer.
When a few percent is the whole argument
Most of the time a word count is a rough sense of length and the disagreement between tools is irrelevant. Three cases where it is not: an essay or submission with a hard limit and a penalty attached, where the marker’s tool is the one that counts; a translation quoted per source word, where a two percent difference on a large job is real money; and a piece written to a publication’s brief, where the editor’s count is the one in the contract.
In all three the fix is the same and it is not to find a more accurate counter, because there is no such thing as the true number of words in a passage — only the number under a stated rule. Quote the tool with the figure. “4,180 words, counted in Word” is checkable; “4,180 words” invites an argument nobody can win.
What this calculator assumes
- The text is written in a language that puts spaces between words. Chinese, Japanese and Thai do not, and every whitespace-splitting counter — this one included — will report one enormous word per line.
- Sentence counts are approximate. A full stop inside an abbreviation, an initial, or a decimal written with a space after it will be read as the end of a sentence.
- Words are compared case-insensitively but not stemmed. “Run” and “running” are two different words in the unique count and in the ranking; so are singular and plural.
- The frequency ranking skips a short list of closed-class English words — the, of, and, to, is and about thirty more — unless you switch that off. It changes only the table. The word count itself always counts every word.
- Characters are UTF-16 code units. Most emoji, and anything else outside the Basic Multilingual Plane, count as two.
- Everything you paste is counted, including headings, captions and references. Academic limits usually exclude some of those, so paste only what the limit covers.
- Nothing is uploaded. The counting happens in your browser — but a share link carries the text in its URL, so do not send one for a confidential draft.
Word count FAQ
Why does this give a different number from Microsoft Word?
Because you are both making judgement calls and you have not made the same ones. The usual culprits are hyphenated compounds, words joined by an em dash with no space around it, numbers written with separators, and whether standalone punctuation counts. Each decision moves the total by a fraction of a percent, and on a 3,000-word document a fraction of a percent is a handful of words. That is normal and it is not a bug in either tool. Where the number is contractual — a translation quote, a competition limit, a client brief — quote the tool alongside the number: “4,180 words, counted in Word” means something; “4,180 words” on its own does not.
Is “don’t” one word or two? What about “well-known”?
Both are one word here. The rule is that a token counts once if, after trimming punctuation off its ends, it still contains a letter or a digit — so the apostrophe inside a contraction and the hyphen inside a compound are interior characters and do not split anything. Two words either side of an em dash are a different matter: this tool treats an em dash, an en dash, a double hyphen and an ellipsis as separators, so “stop—she” is two words. A counter that splits on spaces alone calls it one.
Why is the sentence count labelled approximate?
Sentences are found by looking for full stops, question marks and exclamation marks. That works until a full stop belongs to something else: “Dr. Smith went home.” contains one sentence and two full stops, and this tool reports two. So do most tools. Getting it right needs a dictionary of abbreviations for every language you support, which is a research problem rather than an arithmetic one. On ordinary prose the count is right; on legal, medical or heavily abbreviated text it runs high, and the label says so rather than pretending otherwise.
How many pages is 1,000 words?
About four pages double-spaced or two pages single-spaced, in a 12-point serif face with one-inch margins — the convention that puts roughly 250 words on a double-spaced page. Anything that changes the type size, the leading, the margins or the amount of dialogue changes that, which is exactly why coursework and submission guidelines are written in words rather than pages.
Can I trust the character count against a platform’s limit?
Treat it as close, not exact. Characters here are counted the way JavaScript counts them, as UTF-16 code units, which is the same figure a maxlength attribute enforces — but platforms differ. X counts most characters as one and some scripts as two under its own weighting rules. Google does not truncate a meta description at a character count at all; it truncates by pixel width, so a line of capitals runs out sooner than a line of lowercase. Use the count to get close, then check in the tool that owns the limit.
Sources and review notes
- Unicode Standard Annex #29, Unicode Text Segmentation — the default word and sentence boundary rules, and the standard that says plainly they are approximations needing per-language tailoring
- Unicode Standard Annex #44, the Unicode Character Database — the definitions of the Letter and Number categories this tool tests against when deciding whether a token is a word
Counting a word is a rule rather than a measurement, so there is no data here to go stale. The annexes above are cited for the character categories and the boundary problem, not as a claim that this tool implements their full algorithms — it deliberately does not, and the simpler rules it does use are set out above.