Word and character counter
Words, characters, sentences, paragraphs and reading time, updating as you type. The rules it counts by are written out below, so you can tell why another tool gives a different number.
Words are runs of characters between spaces, ignoring anything with no letter or digit in it. mother-in-law counts as one.
Set a limit above and the room left is counted here.
Most used words| Word | Count | Share |
|---|
Paste some text to see which words you lean on.
What counts as a word here
Every word counter gives a slightly different number for the same text, and none of them explain why. Here is the rule this page uses, so you can decide whether it is the rule you want.
By default a word is a run of characters between spaces that contains at least one letter or digit. That means mother-in-law is one word, don't is one word, 3.14 is one word, and a line that is just — or • is not a word at all. Most word processors work roughly this way, which is why this is the default: if you have been given a 500-word limit by someone looking at a word processor, this is the number that will match them most closely.
Tick Split words the Unicode way and the count comes from your browser's own text segmenter, the same one used for double-click-to-select-a-word. It follows the Unicode segmentation rules, so mother-in-law becomes three words and don't stays one. It is the more linguistically defensible count, and it is usually a little higher.
Chinese, Japanese and Korean
This is where naive word counters fall apart. Chinese and Japanese are written without spaces between words, so a whitespace rule reports a 2,000-character essay as one word. When the text contains CJK characters, an extra CJK characters figure appears, because that is the unit those languages are actually commissioned and measured in — a translation quote or a publishing brief for Chinese counts characters, not words.
If you do want a word count for Chinese or Japanese, switch on the Unicode rule: the segmenter knows the vocabulary well enough to break a sentence into real words rather than individual characters. It is a genuine estimate rather than a guess, though for those languages no two tools will agree exactly, because deciding where a Chinese word ends is a judgement.
Sentences and paragraphs
A sentence ends at a full stop, question mark, exclamation mark or ellipsis that is followed by whitespace or by the end of the text. A run of them — ?! or ... — counts once. Three kinds of false positive are held back: decimal points with a digit either side, single letters before a full stop, and a list of common abbreviations (Mr, Dr, etc, e.g., vs and about twenty more). Text with no terminator at all still counts as one sentence.
The single-letter rule is what keeps J. R. R. Tolkien and U.S.A. from being read as four sentences and three. It has a known cost: a sentence that genuinely ends in one letter, such as “the answer is A.”, is missed. Initials in names are far more common in real prose than single-letter endings, so that is the side to err on, but it is an approximation rather than a rule.
It is a heuristic, as every tool's is. Sentence counts are useful for spotting that your average sentence has crept up to 34 words; they are not useful as an exact figure.
Paragraphs are separated by blank lines. If the text has no blank lines at all, every non-empty line is treated as its own paragraph, which is what you want when you have pasted a list and what you do not want if the rule were applied to prose. Lines are counted separately from paragraphs, so you can see both.
Reading time, and where the number comes from
Reading time uses 238 words a minute, which is the figure from Brysbaert's 2019 meta-analysis of silent reading of English prose across more than a hundred studies. It is deliberately not the 200 or 250 that most sites use as a round number, and it is a mean: actual adult readers spread from roughly 175 to 300 words a minute on ordinary text, and slow down substantially on technical material.
Speaking time uses 130 words a minute, which is a comfortable presentation pace. Audiobook narration sits around 150, conversational speech runs faster, and reading a script aloud at 180 sounds rushed. If you are timing a talk, treat the figure as a floor and add time for pauses, slides and the question you did not expect.
Checking against a limit
Type a limit into the box — 500 words, 160 characters, or just a number, which is read as words — and you get a running count of how much room is left, turning red when you go over. The common ones worth knowing: a meta description is truncated by Google at roughly 160 characters, a single SMS is 160 characters of GSM-7 or 70 if it contains an emoji or an accented character, and a post on X allows 280 characters where most non-Latin characters count double.
Note that character limits are usually counted with spaces, and that platforms differ on what one character means: a count that treats an emoji as one may be talking about code points, while a field that rejects your text is probably counting UTF-16 units or bytes. The figure here counts code points, so a flag emoji made of two regional-indicator symbols counts as two.
The density table
The table lists the words you use most, with each one's share of the total. Its real use is not keyword stuffing — it is catching the crutch word you have used nine times in eight hundred words without noticing. Joining words like the and of are hidden by default because they always win; untick the box to see the raw ranking.
Words are compared in lower case with surrounding punctuation stripped, so Design, and design are the same entry. Nothing is stemmed: design and designing are counted separately, because merging them would quietly hide the repetition you are looking for.
Common questions
Why does this give a different count from Microsoft Word or Google Docs?
Because “word” has no single definition, and every program picks its own rule. The usual points of disagreement are hyphenated compounds, numbers, standalone punctuation, text inside footnotes or headers, and anything not written with spaces. The default rule here — a run between spaces containing at least one letter or digit — is the closest common behaviour to a word processor, and the exact rule is written out above so you can see where a difference would come from.
Does my text get uploaded anywhere?
No. The counting happens in this page, in your browser, as you type. Nothing is sent to a server, nothing is stored and nothing is logged — there is no server involved at all. You can load the page, turn off your network connection, and it keeps working. That matters if the text is an unpublished manuscript, a legal document or client work under an agreement.
Is there a length limit?
Not a fixed one. Counts are recalculated on every keystroke, and a full novel of several hundred thousand words still updates without a visible pause on an ordinary machine. You can also drop a .txt or .md file straight onto the text box instead of pasting.
How are characters counted — with or without spaces?
Both figures are shown. Characters includes spaces, tabs and line breaks; Without spaces excludes all whitespace. Both count Unicode code points, so an accented letter is one character and an emoji is usually one, though emoji built from several joined symbols count as their parts.
Can it count Chinese or Japanese properly?
Yes, in two ways. A CJK characters figure appears automatically when the text contains them, which is the unit those languages are normally measured in. For an actual word count, switch on the Unicode rule and the browser's segmenter will break the text into words rather than reporting the whole paragraph as one.
What is reading time based on?
238 words a minute, from Brysbaert's 2019 meta-analysis of silent reading of English prose. It is an average of adult readers on ordinary material; technical writing is read considerably more slowly. Speaking time uses 130 words a minute, a comfortable pace for a presentation rather than for conversation.
Why does the sentence count look wrong?
Sentence detection is a heuristic in every tool that does it. Decimals, single-letter initials and about twenty common abbreviations are handled, but an unusual abbreviation will add a sentence and a missing full stop will remove one. The single-letter rule also means a sentence ending in one letter is missed. Use the figure to watch your average sentence length move, not as an exact count.
Does it work offline?
Yes. The page is a static file with no network calls after it loads. Once it is open you can disconnect entirely and every count, including the Unicode word rule and the density table, keeps working, because all of it runs on your own device.