Skip to content

How a Word and Character Counter Actually Counts

• text, writing, how-to

A word count looks like the simplest number in writing: you paste the text, a tool says 847, and you get on with the brief. Then a journal says your abstract is 12 words over the limit, a form rejects a caption that “looks like” 80 characters, or a supervisor and a student get two different totals from two different counters. The disagreement is almost never a bug. It is a disagreement about what a word is, and what a character is.

The word counter and the character counter on this site use one set of rules and show them on the page. This article is the longer explanation, so you can decide whether those rules match the brief you were given.

What is a word?

English writers often think of a word as “something with spaces around it”. That works until you hit a contraction, a hyphen, a number, or a language that does not use spaces.

This site counts a word as a run of letters or digits according to the Unicode word-break algorithm, which is the same family of rules a browser uses when you double-click to select a word. A few everyday consequences:

  • don't, it's and won't count as one word each, because the apostrophe is treated as inside the word.
  • state-of-the-art is usually three words, not one. The hyphens are breaks, not glue. If your style guide counts a hyphenated compound as a single word, you will come out slightly high here.
  • Numbers such as 12 or 2026 count as words. They are tokens, not decoration.
  • A Chinese or Japanese sentence, which may have no spaces at all, can still produce a word count, because the algorithm is not “split on spaces”.

No counter will match every house style. Microsoft Word, Google Docs and Overleaf each have their own quirks, especially around footnotes, hyphenation and equations. If a brief says “use Word's count”, use Word's count as the official number, and treat an online counter as a check while you write.

What is a character?

“Character” is an even slipperier word, because computers store text in more than one way.

What a person points to on the screen is a grapheme: one letter, one space, one punctuation mark, or one complete emoji. A family emoji that is stored internally as several code points still looks like a single character, and that is how this site's main character count treats it.

Under the tiles you will also see two other figures:

  • Unicode code points — the number of actual Unicode scalar values. This can be larger than the grapheme count once you use accented letters that are stored as a base letter plus a combining mark, or an emoji sequence.
  • UTF-16 length — what JavaScript's string.length returns, and what some older databases and form limits still use. An emoji often takes two UTF-16 units, so a caption that “looks like 80 characters” can fail an 80-character string.length check.

The main total includes spaces, tabs and line breaks. A second figure strips all whitespace, which is what some “characters without spaces” essay rules ask for. If a form does not say which of these it uses, it is usually the inclusive count, but a developer-written limit may be UTF-16. When you are against a hard cap, confirm in the compose box itself.

Platform-specific limits have further quirks. A short-link wrapper, a CJK typographic rule, or an X compose box may not match this count exactly. The number here is a precise count of the text in the box, not a promise that every destination will agree.

Sentences, paragraphs and lines

Sentences end at a period, question mark, exclamation mark or ellipsis. Common abbreviations such as Dr. are ignored so they do not invent an extra sentence. Initials (J. R. R.) and unusual abbreviations can still throw the number off by one, so treat it as a guide rather than a formal linguistic parse.

Paragraphs are blocks separated by a blank line. If your text has no blank lines — a poem, a list, or a paste from a system that uses a single line break — each non-empty line is counted instead. That is a practical rule, not a claim that every typographic paragraph has been detected.

Lines follow the text as it was pasted, not the wrapping in the box. A trailing newline does not add an empty line, which matches most editors.

Reading time is an estimate, and should be labelled as one

The reading-time label divides the word count by 238 words per minute. That figure is the average silent-reading rate for English non-fiction in Marc Brysbaert's 2019 meta-analysis, published in the Journal of Memory and Language (https://doi.org/10.1016/j.jml.2019.104047). It is a real, cited average — and it is still only an average.

A technical paper, a poem, a text in a second language, or a page full of tables will not be read at that rate. A practised reader of easy prose may be faster. This site shows the time as an estimate and names the rate, rather than presenting “3 min” as a fact about you or about a reviewer. Short texts are shown in seconds so a 40-word caption is not rounded up to “1 min”.

If you need a speaking time, that is a different number again (spoken delivery is typically slower than silent reading). This counter does not invent one.

What this tool will not do

It will not open a PDF or a Word document and promise a reliable count of “the words in the file”. Those formats mix body text with footnotes, headers, alt text and tracked changes; the honest path is to copy or export the text you actually want counted, then paste it. It will not send your essay to a server, and it will not enforce a word limit — it only reports the number.

If you are comparing two drafts rather than counting one, the text diff checker will show what changed. If the words are stuck in a photograph of a page, image to text can lift them first, after which you should still read the result: optical character recognition makes mistakes.

A practical habit

Paste the brief into the box as well as the draft, so you can see the limit and the current total in the same place. Copy the counts if you need to put them in an email to a supervisor or a client. And when a destination has its own official counter, let that one be the official number. An honest online counter is for writing, not for arguing with a portal that has already decided how it counts.

Frequently asked questions

How does this site count a word?

It uses the Unicode word-break algorithm (the same rules a browser uses to select a word on double-click) when the browser supports it. Contractions such as don't count as one word. Hyphenated compounds are usually counted as separate words.

Do spaces count as characters?

Yes, in the main character total, including spaces, tabs and line breaks. A second figure excludes whitespace. Both are grapheme counts, so a typical emoji is 1.

Where does the reading time come from?

Words are divided by 238 per minute, the average silent-reading rate for English non-fiction in Brysbaert's 2019 meta-analysis. It is a planning figure, not a measurement of how fast you will read this particular text.

Is the text uploaded?

No. Counting runs in your browser. You can paste or load a text file; nothing is sent to a server.