Why Word Counts Differ Between Tools

The short answer

That rule is simple and predictable, which is its virtue. It also means state-of-the-art is one word, 1,250 is one word, and a web address is one word. A tool that splits on punctuation instead would give you four, two and several.

Where counters disagree

Hyphenated and compound words

"A state-of-the-art, well-known result" is 4 words and 37 characters under a whitespace rule. Replace the hyphens with spaces — "A state of the art, well known result" — and it becomes 8 words, while the character count stays at 37. Nothing changed except punctuation, and the word total doubled. Publishers that pay by the word usually say so in their style guide for exactly this reason.

Numbers, dates and units

"We shipped 1,250 units on 3 May 2026 at 4.5 kg each." is 12 words and 52 characters. The whitespace rule counts 1,250 and 4.5 as single words, and counts 3, May and 2026 as three separate words. A counter that ignored numeric tokens entirely would report 9; one that split 1,250 at the comma would report 13.

Contractions and apostrophes

"It isn't ready, so we'll wait." is 6 words and 30 characters. Whitespace splitting keeps isn't and we'll whole. Expanding them to "is not" and "we will" would give 8 — a real difference when a limit is tight. Curly and straight apostrophes make no difference to the word total, but they are different characters, so find-and-replace work on them is not interchangeable.

URLs, file paths and code

"See https://example.com/docs/page?id=12 for details." is 4 words and 52 characters: the entire address is one word. Most of the length is in that one token. If you are writing to a character limit, a long link can consume a quarter of your allowance while barely moving the word count.

Spacing itself

Two spaces between sentences do not change the word count, because the splitter treats any run of whitespace as a single break — but they do change the character count. "one two" with a double space is 2 words and 8 characters; with one space it is 2 words and 7. Line breaks and tabs behave the same way: they separate words and each counts as one character. Leading and trailing spaces are trimmed before words are counted, so " hello " is 1 word, but they are still counted among the 9 characters.

Characters: code units, code points and what you see

"Characters" has three plausible meanings, and they agree only for plain unaccented English.

• Code units — the storage slots JavaScript uses. Most characters take one; anything outside the basic range, including most emoji, takes two.
• Code points — one entry in the Unicode catalogue. An emoji is usually one code point; a letter plus a separate combining accent is two.
• Grapheme clusters — what a reader would call a character, including an accented letter written as a base plus a mark, and a multi-part emoji.

The family emoji is the clearest case. It is one picture on screen, but it is three person emoji joined by two invisible zero-width joiners: 5 code points stored in 8 code units. A platform that limits a message to 280 characters has to pick one of these measures, and the one it picks decides whether that single emoji costs you one character or eight.

Combining accents cause the quieter version of the same problem. Two visually identical words, café and café, can be 4 or 5 characters long depending on how the text was typed or pasted. Normalising text before counting removes the discrepancy; few counters do it.

With spaces or without

Typesetting and translation work often quotes characters without spaces, which simply means whitespace is stripped before measuring. "Nice work 😀" is 12 characters with spaces and 10 without. Neither figure is more correct; they answer different questions, so check which one a brief is asking for.

How this site's counter decides

1. The text is trimmed of leading and trailing whitespace. Empty or whitespace-only text reports 0 words.
2. What remains is split on every run of whitespace — spaces, tabs, newlines — and the number of pieces is the word count. Punctuation, hyphens, apostrophes and digits never split a word.
3. The character figure is the length of the untrimmed text in UTF-16 code units, so it includes every space, tab and line break, and counts most emoji as two.
4. The result is printed as a single line, for example Words: 4 | Characters: 37.

Everything happens in your browser; the text is never sent anywhere. One consequence of step 2 worth remembering: a stray punctuation mark surrounded by spaces, as in "cost - benefit", counts as a word of its own, making that phrase 3 words rather than 2.

Common mistakes

• Assuming a word processor will agree. Office software usually counts footnotes, headers, captions and text boxes under its own rules. Paste the body text alone if you need a comparable figure.
• Writing to a limit in the wrong unit. Confirm whether a 500-character limit means code units, visible characters, or characters without spaces before you trim.
• Counting formatted text. Markdown symbols, HTML tags and tracked-change markup all count as characters if you paste them in. Strip them first.
• Expecting a readability score to match. A reading-ease measure has to divide the text into sentences as well as words, so an abbreviation such as "e.g." can be read as a sentence boundary and move the score without changing the word count at all.
• Treating emoji and accents as one character each. They frequently are not, and a form that validates length server-side may reject text your counter said was short enough.
• Copying from a PDF and trusting the result. PDF text extraction often inserts line breaks mid-sentence or drops spaces, which changes both totals.

Check it yourself

Paste any of the examples above into the word and character counter and press the count button. "A state-of-the-art, well-known result" returns Words: 4 | Characters: 37; the hyphen-free version returns 8 words and the same 37 characters. The counter also clears, copies and downloads the box as a plain text file, which is a convenient way to keep a draft you were counting.

For the sentence-level view, run the same text through the Flesch reading ease checker and watch what happens when you split one long sentence into two: the word count is unchanged, the score moves. That contrast is the clearest demonstration that a count is only as meaningful as the rule behind it.

Related tools and guides

• Word and character counter
• Flesch reading ease checker
• Text summarizer
• Image quality vs file size

Reviewed October 1, 2026. Calculation methods and corrections.