Word Count Rules: Why Two Tools Disagree About the Same Text

Most counters split on whitespace, which makes "state-of-the-art" one word. A counter that splits on punctuation makes it four. Contractions usually count as one, numbers usually count as one regardless of length, and a standalone symbol may or may not count at all. A date written as 1 January 2026 is three words while 2026-01-01 is one. None of these choices is incorrect; they are simply different conventions, and the gap between two tools on a long document is routinely one or two percent.

Most institutions exclude the reference list, the title page and any appendices from a word limit, and most include the body text and quotations. Footnotes, abstracts, tables and figure captions vary by department and are the usual source of disputes. In-text citations are normally counted, which means a citation-dense paragraph consumes a meaningful share of the limit. Because these rules are set locally rather than universally, the handbook governs, and a general-purpose word counter cannot apply the exclusions for you.

Publishers and translators frequently work in characters rather than words, and the two figures differ by roughly the word count itself, since one space follows most words. Translation is often priced per character or per thousand characters, subtitling is constrained by characters per line, and form fields usually enforce characters including spaces. Check which measure the constraint uses before editing to fit it, because trimming for a word target and trimming for a character target lead to different edits.

The common rule of 500 words per double-spaced page assumes a 12-point serif face, one-inch margins and no headings, tables or figures. Change any of those and the count moves substantially. A page of dialogue with frequent line breaks holds far fewer words than a page of unbroken prose. When a requirement is expressed in pages, the safest reading is to ask for the underlying formatting, or to submit against the word figure if one is offered.

Chinese and Japanese are written without spaces between words, so whitespace splitting collapses an entire sentence into one token. Counts for these languages are therefore based on characters, and a rough working equivalence is that one English word corresponds to roughly 1.5 to 2 Chinese characters of comparable content. Mixed-language documents need a counter that segments each script separately, otherwise an English abstract attached to a Chinese body produces a total that describes neither.

Reading time is normally word count divided by an assumed rate, commonly 200 to 250 words per minute for silent reading of general prose. That rate is far too optimistic for technical material with code samples, and far too slow for skimmed listicles. Treat the figure as a rough signal of length rather than a measurement, and if it is displayed to readers, keep the assumed rate consistent across a site so the numbers stay comparable.

Decide which convention you need before you start editing. If the limit is academic, remove the reference list and any explicitly excluded sections, then count the remaining body. If the limit is a form field, switch to characters including spaces. If the text is mixed Chinese and English, use a counter that reports both word and character totals. Re-count after the final edit rather than trusting an intermediate figure, because a single pasted citation block can move the total by dozens of words.