Word counting sounds like it should have exactly one correct answer for any given piece of text — count the words, report the number. In practice, different word processors, writing platforms, and online counters routinely disagree on the count for the exact same text, sometimes by a meaningful margin, because "what counts as a word" turns out to have several genuinely different reasonable definitions.
The basic method: splitting on whitespace
The simplest and most common approach counts words by splitting text wherever there's whitespace and counting the resulting chunks. This is fast, easy to implement, and matches most people's intuitive sense of what a word count should measure. But it has an immediate edge case: what happens with multiple consecutive spaces, or a stray tab character? A naive implementation that doesn't account for this can produce phantom empty "words" from consecutive whitespace, inflating the count above what a careful manual count would produce.
Hyphenated words: one word or two?
A compound word like "well-known" or "twenty-three" is visually one connected unit, but a whitespace-splitting approach will correctly count it as a single word since there's no space breaking it apart — good. But a hyphenated phrase used more like a parenthetical aside, such as "the meeting—rescheduled twice already—finally happened," uses a similar-looking dash character in a way that's grammatically closer to a sentence break than a compound word, and different counting tools handle this distinction differently depending on how carefully they were built to distinguish a genuine hyphen from an em dash or en dash used for a different grammatical purpose.
Numbers, symbols, and standalone punctuation
Is "42" a word? Most word counters say yes, since it's a standalone token separated by whitespace, even though it's not technically a word in the linguistic sense. What about a standalone em dash, or an ellipsis used as its own separated token? These edge cases rarely come up in typical prose, but they do show up in more stylized or informal writing, and different tools make different implicit decisions about whether to count them, which is part of why the same creatively-punctuated text can produce different counts across different counters.
Contractions split differently depending on the apostrophe character used
"Don't" is clearly one word to any human reader, but the specific apostrophe character matters more than you'd expect for an automated counter — a straight apostrophe (') versus a curly, typographic apostrophe (') are different Unicode characters, and a counting algorithm that only recognizes one specific apostrophe character as part of a word (rather than a word-breaking character) will incorrectly split "don't" into two separate word-like tokens if the "wrong" apostrophe variant is used. Text copied from different sources often mixes these two apostrophe styles inconsistently, which is a surprisingly common source of word count discrepancies between two copies of what looks like identical text.
Why platform-specific word limits sometimes feel wrong
This matters practically whenever you're working against a strict word limit — a submission requirement, an assignment cap, a platform's character or word restriction — because the count your own writing tool reports and the count the receiving platform enforces might use subtly different counting rules, especially around hyphenation, numbers, and punctuation-heavy text. A piece of writing that shows as exactly at the limit in one tool can show as slightly over or under in another, purely due to these differing edge-case decisions rather than any actual difference in the underlying text.
There isn't a single "correct" answer, only more or less common conventions
None of the counting approaches described above is objectively wrong — they're each internally consistent, reasonable interpretations of an inherently slightly ambiguous question. The practical implication is that if a specific platform enforces a strict word limit, it's worth checking that same platform's own word counter (if it provides one) rather than assuming any other tool's count will match exactly, since the small edge-case differences described here can occasionally push a borderline piece of text from just-under to just-over a strict limit.
Counting your own text
Our word counter uses a straightforward whitespace-based approach that matches how most word processors and writing platforms count in the common case, giving a reliable reference figure — just keep in mind, per everything above, that a strict external word limit enforced by some other specific platform might count a handful of genuine edge cases slightly differently.