What Counts as a Word — and Why Counters Disagree

Every word counter has a definition of a word, and very few of them state it. This guide states the definition behind the WordCntr.com word counter rule by rule, runs short original sentences through the same engine the counter uses, and shows exactly where two honest counters can disagree. Every number on this page is computed at build time by that engine and interpolated into the text, so if the rule ever changed, the examples would change with it.

The short version: a word is an unbroken run of letters, digits, apostrophes, and hyphens. Everything interesting on this page follows from what that sentence includes and what it leaves out.

The rule, in one sentence

A word is an unbroken run of Unicode letters, digits, apostrophes (straight or curly), and hyphens. Everything else separates: spaces, line breaks, periods, commas, colons, slashes, quotation marks, em dashes, emoji, currency symbols. There is no dictionary lookup, no minimum length, and no attempt to understand the text: each maximal run of those characters is one word-token, and the word count is the number of tokens. "Letters" means any script Unicode classifies as letters, not only A–Z, so accented and non-Latin text counts correctly; the letters guide goes deeper on that.

One paragraph, token by token

The methodology page uses one paragraph to show every counter at once: "Dr. Smith arrived at 9 a.m. She left early." The engine reports 10 words, 43 characters (35 without spaces), 3 sentences, and 1 paragraph. The word-tokens are "Dr", "Smith", "arrived", "at", "9", "a", "m", "She", "left", and "early". Three things in that list are worth pausing on. "Dr" loses its period, because a period is not a word character and simply ends the run. "9" is a token in its own right, because digits are word characters. And "a.m." becomes "a" and "m" — two tokens — because the periods inside it separate exactly as the space after it does. The engine does not know that "a.m." is an abbreviation; it knows only which characters join and which split.

The sentence figure of 3 is higher than the answer a reader would give, because the period after "Dr" is followed by a space and so triggers the sentence rule. That over-count is a documented limitation of the sentence counter, explained in the sentences, paragraphs, and lines guide; it leaves the word count untouched.

Hyphens and apostrophes join

Because hyphens and apostrophes are word characters, a compound or a contraction stays whole. Take "The well-known café's state-of-the-art heater isn't ready.": the engine reports 7 words — "The", "well-known", "café's", "state-of-the-art", "heater", "isn't", and "ready". The accented é is a letter like any other, so "café's" is a single token from c to s. Retype the sentence with curly apostrophes — "The well-known café’s state-of-the-art heater isn’t ready." — and the count is 7, unchanged. Whether an editor inserts straight or curly apostrophes makes no difference to the count, because both forms are in the rule.

This is the boundary where counters most often part ways. A counter that treats the hyphen as a separator sees the same sentence as 11 words: "well", "known", "state", "of", "the", and "art" become tokens of their own. The gap between the two answers is 4 — one extra token per hyphen — and it is a convention, not a bug in either tool. The case converter's Title Case uses the same token, which is why "state-of-the-art" receives exactly one capital there (see the case conversion guide).

Digits are word characters; separators are not

Digits join runs exactly as letters do, so "2026" is 1 word. What the rule does not do is recognize the punctuation that humans read as part of a number. A decimal point is a period, and a period separates: "3.14" is 2 words, "3" and "14". A thousands separator is a comma, and a comma separates: "3,140" is 2 words, "3" and "140". Run the sentence "Build 2026 measured 3.14 seconds, then 3,140 more." through the counter and it reports 10 words, where a reader counting the figures as single items would arrive at fewer. The tokens are "Build", "2026", "measured", "3", "14", "seconds", "then", "3", "140", and "more".

The consequence is directional: a passage dense with decimals, prices, or large figures counts higher here than a reader would tally it, by one token per internal period or comma. The sentence counter, by contrast, handles the decimal correctly — "Pi is 3.14 exactly." is one sentence — because its rule requires whitespace after the period, and the period inside "3.14" has none.

What a pasted URL does to the count

Colons, slashes, and dots all separate, so a web address fragments into several tokens. On its own, https://example.org/notes/word-rules is 5 word-tokens: "https", "example", "org", "notes", and "word-rules" — the last of them held together by its hyphen. In a sentence, "Read https://example.org/notes/word-rules before Friday." counts as 8 words, where a reader would probably count the link as one word or skip it altogether. An email address behaves the same way, splitting at the @ and at each dot. A counter that treats the address as a single word would report 4 fewer for that sentence — all but one of the tokens the address contributes here — so if a limit matters, remove links before counting or find out how the recipient's tool treats them.

The bare-hyphen quirk

The rule has one consequence no reader would endorse, documented rather than hidden. Since hyphens and apostrophes are word characters and there is no minimum length, a hyphen standing completely alone — whitespace on both sides — registers as a token. A plain-text list of three lines, "- apples", "- pears", "- plums", counts as 6 words; the same three items without the leading hyphens count as 3. The difference is exactly the number of bare hyphens. A lone straight apostrophe behaves the same way. An em dash does not: "apples — pears — plums" is 3 words, because the em dash is not in the rule and simply separates.

If you draft lists with hyphen bullets and the count matters, subtract one per bullet or switch to an asterisk, which is not a word character. How the same list counts as lines and paragraphs is covered in the sentences guide.

Unique words are compared case-insensitively

The unique-word figure the engine reports beside the word count lowercases every token before comparing, and nothing more. "The tide turns; the Tide always turns." is 7 words — "The", "tide", "turns", "the", "Tide", "always", "turns" — and 4 unique words, because "The" and "the" collapse into one entry, as do "tide" and "Tide", as do the two copies of "turns". Lowercasing is the only merge applied: there is no stemming, so "turn" and "turns" would stay separate, and no apostrophe stripping, so "café's" and "cafés" would too.

Why counters disagree

There is no standard definition of a word for counting purposes. Every counter chooses a rule set. Where another counter uses a similar rule, ordinary prose — sentences of letters, spaces, and end punctuation — will agree closely, because nothing in such text is a boundary case under either rule. The disagreements cluster at the boundaries this page has walked through: hyphens, apostrophes, punctuation inside numbers, web and email addresses, and stray punctuation marks. A reference list, a spreadsheet paste, or a paragraph full of hyphenated compounds will diverge between tools by a predictable amount, in a predictable direction. The sentence "The well-known café's state-of-the-art heater isn't ready." is 7 words under this site's rule and 11 under a hyphen-splitting one, and both tools are doing what they say they do.

The other counts behave the same way: some platforms count characters as UTF-16 code units rather than code points, word processors count paragraph marks rather than blank-line-separated blocks, and abbreviations inflate any sentence count built on end punctuation. Each is a convention with a stated consequence; the characters guide covers the first in detail.

For a hard limit, the recipient's counter wins

When a submission carries a word limit, the count that matters is the one produced by whoever enforces it — the form, the platform, the reviewer's software — because that rule set, not this site's, decides whether the text is accepted. This page's count is the honest baseline, and the sections above tell you which direction another rule would push it: a tool that splits hyphens counts higher, by one per hyphen; a tool that treats a link as a single word counts lower, by all but one of the tokens the link contributes here; a tool that ignores bare punctuation never sees the bullet hyphens.

The working practice is simple. Draft against the word counter, and when the text is within a few words of the limit, paste the final version into the recipient's own tool and trust its number over any other. If the two disagree, look for hyphens, figures, links, and bullets before assuming either tool is wrong. The reading time guide shows how the word count feeds the estimates built on it; the guides index covers the rest.

Frequently asked questions

Counting happens locally in your browser — nothing you type is transmitted or stored. Every number in this guide is computed at build time by the same engine the word counter runs; conventions are named where they apply. See the methodology page.