Counting words looks like it should have a single correct answer, and the reason it does not is the most interesting thing about this tool. A counter has to decide what a word is, and every reasonable definition disagrees with the others. Does a hyphenated compound count once or twice? Does a number count? Is a Chinese character a word, when Chinese has no spaces at all? These are not pedantic questions when a limit is attached to the answer. A meta description has a length before search engines truncate it, a journal has a word limit, a submission portal counts in its own way and rejects a paper that is four words over, and a caption has a character budget. If two counters disagree by a hundred words on the same text, the usual explanation is not a bug in either of them - it is a different definition of the same word, and the useful move is to find out which rule the party imposing the limit actually uses.
Why two counters disagree about the same paragraph
The classic difference is the hyphen. Microsoft Word and Google Docs treat a hyphen, a slash or an em dash as a word boundary, so well-known is two words and a well-known state-of-the-art system is four, while a simple counter that splits on whitespace calls it one. A rule built from a regular expression such as a word character class disagrees with both: it counts dont as two words, because the apostrophe is not a word character, and it counts 3.5 as two. A standalone number is generally one word in Word and one in most counters, which is why a table full of figures can inflate one tool's total and not another's. Where the numbers diverge most in practice is pasted text: copy a paragraph out of a document or a slide and the straight quotes and non-breaking spaces come with it, and different tools treat those differently. If your count is close to a limit, re-check with the counter belonging to the system that will judge the submission.
Counting Chinese, Japanese and Korean text
None of these languages separates words with spaces, so a word counter has to choose a convention, and the two common ones give very different totals. The first counts every Han character as one word, which matches how Chinese word processors display the character total and how most submission portals count a Chinese abstract. The second treats a run of characters between punctuation marks as one word, which is closer to the linguistic idea of a word and produces a noticeably smaller number. Japanese adds a wrinkle because particles and inflection are written without spaces and are arguably not words on their own, and Korean places spaces between grammatical units, which are longer than English words. Practical advice for anything written in Chinese: quote the character count, not the word count, and check whether the target you are submitting to counts characters or words, because the same form often wants one and a school essay wants the other. Latin words and digits embedded in Chinese text are usually counted as one word each on top of the characters.
Sentences, paragraphs and reading time are estimates too
Sentences are counted by looking for a full stop, a question mark or an exclamation mark, and that method breaks on the text it is most likely to see: abbreviations such as e.g. and Mr., version numbers like 2.0, decimal numbers, ellipses, and a final period followed by a closing quote. Expect the sentence count to be an overcount on academic or technical writing, sometimes by ten percent. Paragraphs are counted as blocks of text separated by a blank line in most tools, but a document that uses single newlines will report one paragraph where an editor would report three, and pasted text often loses the blank lines entirely. Reading time is derived, not measured: it multiplies the word count by an assumed reading rate, typically somewhere around 200 to 250 words per minute for adult prose, and real readers sit well below that on technical material and above it when skimming. Treat a reading time as a range, not a promise.
Where a character count actually bites
Limits are everywhere and they are rarely round numbers. Search engines truncate a meta description somewhere around 150 to 160 characters on a desktop result and closer to 120 on a mobile one, so a description written to the limit will often be cut mid-sentence on a phone. Title tags are usually treated as roughly 50 to 60 characters before they are rewritten in the result. Alt text has no exact number, but a screen reader user benefits from a short sentence rather than a paragraph. Social posts have hard caps in characters, submission forms have limits in words or characters that vary by field, and column widths on a print layout are measured in characters per line, which is one of the few uses of a character count that is genuinely predictive. Counting the raw length of a string in JavaScript also deserves a footnote: the length property counts UTF-16 code units, so an emoji outside the basic plane counts as two and an ordinary Chinese character counts as one.
Frequently asked questions
Does this send my text anywhere?
No. Counting happens in your tab as you type, and nothing is transmitted. That matters for the thing people most often count: an unpublished draft, a manuscript, something a client has not seen yet. The site is supported by advertising, so ad providers can see page-level information such as your IP address, but not the contents of the editor.
Why does my count differ from Microsoft Word?
Almost always the definition of a word. Word splits on hyphens, slashes and dashes, so well-known is two words there and one word to a whitespace splitter. Word also treats a standalone number as one word and handles apostrophes its own way. Copy the text into plain form, or accept that the two tools are answering slightly different questions.
How is reading time calculated?
By multiplying the word count by an assumed reading rate, usually around 200 to 250 words per minute. That is an average for adult prose, and it does not fit technical documents, dense reference material, non-native readers or anyone skimming. Use it as a rough indicator, and if you need a real number, read the piece and time yourself.
Is it free and does it need an account?
Free, with no sign-up and no daily limit on what a single document can be. Once the page is loaded the counting is local, so it keeps working if the connection drops, and everything is cleared when you close the tab.