Word Counter vs MS Word Count: When Browser Wins

You paste a 750-word draft into an online counter and it says 748. Word says 751. Neither is broken. Word count is not a single defined number; it is whatever a tool's tokenizer decides a "word" is, and different tools draw the boundaries in different places. This piece explains exactly where a browser word counter and MS Word's word count diverge, why the gap is usually harmless, and the cases where the browser is the tool you actually want.

#TL;DR

• Word counts vary because each tool tokenizes differently — hyphens, numbers, and URLs are the usual culprits.
• MS Word counts characters both with and without spaces; confirm which one a limit means.
• Word counts CJK text by characters, not spaces, because those scripts have no inter-word spacing.
• A browser counter runs live, needs no install, and never uploads your text.
• Use Word for tracked, footnoted manuscripts; use the browser for quick, private, cross-format counts.

#How does Microsoft Word count words?

Word has counted words as you type since long before the web had a good answer, and its numbers show in the status bar plus a detailed dialog. Per Microsoft's own Show word count documentation, the count covers words, pages, paragraphs, lines, and characters. The rules behind that single "words" figure are where the surprises live.

#What counts as a word in Word

Word's tokenizer is looser than a naive "split on spaces". A hyphenated compound like state-of-the-art counts as one word, not four. A formatted number like 1,000 counts as one word, and so does 3.14. An email address or URL usually counts as one token too. This is sensible for prose but means Word's number reflects editorial units, not whitespace-separated chunks.

#Characters with spaces vs without spaces

Open Review → Word Count and you get two character figures: characters with spaces and characters without spaces. The distinction matters more than people expect. Social platforms and most CMS limits count characters with spaces — every keystroke, blanks included. Many academic submission portals and translation billing systems, especially for Asian language pairs, count characters without spaces. A 280-character tweet limit and a 280-character abstract limit can refer to two different measurements.

#Why footnotes and text boxes change the number

By default, Word's status-bar figure excludes text inside footnotes, endnotes, and text boxes. The Word Count dialog has a checkbox to include them, and toggling it can move a manuscript's total by hundreds of words. If a colleague's count disagrees with yours by a suspiciously round amount, the footnote checkbox is the first thing to check.

#Why does a browser word counter give a different number?

A browser tool has no legacy document model to honor, so it can pick a modern, transparent tokenizer. That freedom is exactly why its number can differ from Word's — and often why it is more predictable.

#How online word counters segment text

Most counters treat a "word" as a run of characters separated by whitespace, then optionally refine that with Unicode-aware rules. The authoritative reference for where a word begins and ends is the Unicode Consortium's UAX #29: Unicode Text Segmentation, which defines default word boundaries for every script. A quality browser counter follows those rules rather than inventing its own, which makes its behavior documentable instead of folkloric.

#Intl.Segmenter and word-like tokens

Modern browsers ship a native implementation of UAX #29 in Intl.Segmenter. It returns each segment with an isWordLike flag, so a counter can include real words and skip punctuation and spaces. Per MDN's Intl.Segmenter documentation, the API is locale-sensitive, which is what makes it reliable across scripts that a space-splitter mangles.

function countWords(text, locale = "en") {
  const seg = new Intl.Segmenter(locale, {
    granularity: "word",
  });
  let n = 0;
  for (const s of seg.segment(text)) {
    if (s.isWordLike) n++;
  }
  return n;
}

countWords("state-of-the-art design"); // 4

Note that Intl.Segmenter splits state-of-the-art into four word-like tokens, whereas Word treats it as one. Same text, different-but-defensible answers.

#The hyphen, URL, and number edge cases

The biggest gaps between any two counters trace back to a short list of tokens:

• Hyphenated compounds: one word (Word) vs several (space/Unicode splitters).
• Numbers with separators: 1,000 and 3.14 may be one token or two.
• URLs and emails: one token in Word; a naive splitter keeps them whole, a Unicode splitter may break on punctuation.
• Em-dashes with no spaces: word—word can read as one token or two.
• Emoji and ZWJ sequences: character counts diverge sharply; see the note on CJK and emoji below.

#Word Counter vs MS Word: a side-by-side

The table below shows the same short inputs measured two ways: MS Word's tokenizer versus a Unicode word-boundary counter like the one in a modern browser.

Neither column is wrong. They answer slightly different questions: Word reports editorial words, the browser reports Unicode word boundaries.

#Where the two agree

For ordinary running prose — sentences of unhyphenated English words separated by single spaces — the two approaches agree almost perfectly. A 2,000-word article will typically match within a handful of words, and that drift comes entirely from the edge cases above, not from either tool miscounting.

#Where they diverge

Divergence scales with how "technical" your text is. Legal writing full of hyphenated terms, developer docs full of URLs and code identifiers, and financial copy full of formatted numbers will show the widest gaps. If your count has to match a specific system, count in that system.

#How to count Chinese, Japanese, and Korean words correctly

CJK text is where naive counting fails hardest, and where knowing your tool matters most.

#Why whitespace splitting undercounts CJK

Chinese, Japanese, and Korean do not put spaces between words. A splitter that keys on whitespace sees 中文单词是没有空格的 as a single token and reports "1 word" for a whole sentence. That is why, for CJK, character count — not word count — is the dependable billing and length metric. UAX #29 addresses this directly: it places word boundaries around each ideograph, so a Unicode-aware counter can report meaningful segments instead of one giant blob.

#What Word does for Asian text

Word handles this by counting CJK characters as the primary unit. Its "Asian characters, Korean words" line gives the ideograph count and excludes Latin words and digits. A browser counter using Intl.Segmenter with an appropriate locale reaches the same conclusion through the Unicode rules:

const zh = new Intl.Segmenter("zh", {
  granularity: "word",
});
let words = 0;
for (const s of zh.segment("中文单词")) {
  if (s.isWordLike) words++;
}
// words === 4  (one per ideograph)

If you routinely work across scripts, our deeper dive on how to count Chinese, Japanese, and Korean words walks through the billing and localisation implications.

#When the browser beats the desktop

MS Word is excellent at what it is: a full document processor. But counting words is a tiny job, and for that tiny job a browser tool wins more often than habit suggests.

#Speed, privacy, and no install

A browser word counter opens in a tab, updates live as you paste or type, and — crucially — never sends your text anywhere. Everything runs in the page. For a confidential draft, a client's unreleased copy, or code comments you would rather not upload, in-browser counting is a genuine privacy win, not a marketing line. There is nothing to install, nothing to license, and it works the same on a locked-down work laptop as on your phone. It also happily counts text you would never open in Word: a chunk of Markdown from our Markdown editor, or the differing halves of two drafts you are comparing in the diff checker.

#When you should still use Word

Reach for Word when the count is tied to the document itself: manuscripts with tracked changes, academic papers where the footnote-inclusion rule is defined by the journal, or long files where you need per-section counts by selecting text. Word's ability to count a selection and include or exclude notes on demand is hard to beat inside a live editing session. The honest rule: count where the limit is defined. If your target lives in Word, count in Word; for everything quick, private, or cross-format, the browser is faster and safer.

#References

• Show word count — Microsoft Support — official description of what Word's word, character, line, and paragraph counts include.
• UAX #29: Unicode Text Segmentation — the standard defining default word boundaries across scripts, including ideographic text.
• Intl.Segmenter.prototype.segment() — MDN — browser API used for the word-counting code samples and the isWordLike flag.

#Related on iKit

• See live word, character, and reading-time counts as you type — the base guide to the counter this comparison is built around.
• Count Chinese, Japanese, and Korean text the right way — why CJK needs character counts, expanding on the divergence covered here.
• Why str.length lies about emoji length in JavaScript — the character-counting counterpart to this word-counting piece.
• Fit SEO titles and meta descriptions to real length limits — where the with-spaces vs without-spaces distinction bites in practice.