Counting words sounds like the simplest thing a computer could do, and it is not. Where a word ends is a question about language, not about spaces, and every counter that disagrees with another one is disagreeing about that. Here is what this one counts and where it is honest about giving up.
Splitting on spaces is wrong for most of the world
The obvious implementation — cut the text wherever there is whitespace — works for English, Italian, German and every other language that puts gaps between words. It fails completely for the ones that do not. Chinese, Japanese and Thai are written without spaces, so the naive approach reports a full paragraph as one word.
This tool uses Intl.Segmenter, the browser's built-in text segmentation, which knows the rules for each language rather than looking for gaps. 我喜欢吃苹果和香蕉 counts as 6 words rather than 1, and 私は日本語を話します as 7. For English the answer is identical to the naive one, so nothing changes for most visitors — it just stops being wrong for everyone else.
Segmentation is genuinely hard and even a good segmenter is making judgements. Whether "New York" is one word or two, whether a hyphenated compound counts once or twice, whether an apostrophe splits a token — these have no universal answer, which is the real reason two counters disagree.
A character is not what JavaScript thinks it is
JavaScript strings are sequences of UTF-16 code units, and a code unit is not a character. Ask JavaScript for the length of a thumbs-up emoji and it says 2. Ask for a family emoji — the kind built from several people joined together — and it says 8. Neither number is what any human would call a character count.
This tool counts grapheme clusters instead: the units a reader would point at and call "one character". The family emoji is one. An accented letter is one whether it was typed as a single code point or composed from a letter plus a combining mark. If you are comparing this count against a limit imposed by some other system, be aware that the other system is quite likely still counting code units, and will disagree with you the moment an emoji appears.
Sentence counting is a heuristic and it is wrong sometimes
This is the part worth being upfront about. The tool counts sentences by looking for full stops, question marks and exclamation marks. That is a heuristic, and heuristics fail in predictable ways.
"Dr. Smith arrived." counts as two sentences, because the abbreviation carries a full stop. "Pi is about 3.14 exactly." counts as two, for the same reason. Ellipses do behave correctly — "Wait... really?" is two, because runs of punctuation collapse into one match — but that is luck rather than design.
Doing better means detecting abbreviations, decimals, initials, URLs and quotations, which is a language-specific problem with no small solution. The number here is a useful approximation for prose, and you should not trust it on anything technical. Knowing the limitation is more useful than a number that pretends it has none.
The reading time is an average and you are not average
The figure comes from dividing the word count by 200, which is a widely used convention for adult reading speed in English. Actual silent reading speeds range from about 200 to 300 words per minute for adults reading prose, and the spread between individuals is far wider than that.
The number also assumes the text is prose. Technical documentation, code, dense reference material and anything you re-read are all much slower. Treat it as a rough signal — "this is a five-minute read, not a thirty-second one" — and not as a measurement. It has become a standard on blogs because it sets expectations, not because it is accurate.
Common questions
Is my text sent to a server?
No. This tool is marked "client": everything is counted in your browser tab as you type. Nothing is transmitted and nothing is stored, which matters here because word counters get pasted with drafts, essays and manuscripts people would rather not hand to a stranger.
Why does my word count differ from Microsoft Word?
Because you have asked two programs the same ambiguous question. Word has its own rules about hyphens, numbers, URLs and what counts as a word at all, and those rules are not the same as the browser's. Neither is wrong. If you have a hard limit to hit, count with whatever tool the person imposing the limit will use — that is the only count that matters.
Does the character count include spaces?
Both are shown. The distinction matters more than it looks: publishers and typesetters usually mean "with spaces", while character limits in forms and databases usually mean every character including them. When someone gives you a limit without saying which, they nearly always mean with spaces.
Why does an emoji count as one character here but two elsewhere?
Because most systems count UTF-16 code units and most emoji occupy two of them. This tool counts what you would see. The disagreement is real and matters for anything with a hard limit — a database column defined as 100 characters may reject text this tool says is 95, because it is counting differently. If a limit is being enforced somewhere, find out what that somewhere counts.
Can I trust the sentence count for academic work?
Not without checking it. Any text with abbreviations, initials, decimals or citations will be over-counted, and academic writing is full of all four. For a rough sense of sentence length across a draft it is fine. For anything you are going to report as a figure, count them yourself.