A word count answers a different question in Chinese than it does in English. English is written with spaces, so counting "words" means counting the tokens between separators; Chinese is written without spaces, so the smallest unit a tool can count without a dictionary is the character. Because one Chinese character carries roughly the information of an English word, the counting conventions — and the reading speeds that follow from them — differ: about 200 words per minute for English and 400 characters per minute for Chinese.
That difference is not academic. It decides whether a 500-word commission becomes a 500-character one, whether a subtitle cue fits its two seconds on screen, and whether a translation quote is priced per word or per character.
What a "word" is in each writing system
In English, the boundary is visible: spaces, hyphens and apostrophes tell the reader where words end. Even so, the definition is not perfectly clean. Ice cream is two tokens and one concept. Cannot is one token and two words. Don't is one token with an apostrophe, and depending on the tool it may be counted as one word or two. Most counters treat any run of letters, digits and internal apostrophes or hyphens as a single word, which is why state-of-the-art usually counts once.
Chinese has no such visible boundary. A sentence like 我们今天开会 runs together as five characters, while the "words" a linguist would identify are 我们 (we), 今天 (today) and 开会 (to hold a meeting) — three words, five characters. Splitting that string into words requires a segmentation dictionary and a set of rules for ambiguous cases, and different segmenters disagree at the margins. Counting characters requires none of that: every character is one code point, and the total is identical no matter which tool you use.
Japanese mixes the two problems. Kanji, hiragana and katakana are written without spaces, so the manuscript convention is characters — the traditional genkō yōshi paper holds 400 characters per page, and assignments are still specified in 400-character pages. Korean sits closer to English: eojeol are separated by spaces, so a Korean word count is closer to a space-delimited token count, though each eojeol can bundle a stem with several particles.
This is why serious counters report more than one number. A single "word count" for a bilingual document is a fiction; characters, space-delimited tokens and CJK characters are three different measurements of the same file.
Why Chinese is counted at 400 characters per minute
Reading speed is usually expressed as words per minute in English because words are the unit readers fixate on. For Chinese, the comparable unit is the character: adult readers of Chinese average roughly 300 to 500 characters per minute for general non-fiction, and 400 is the round number that most Chinese-language publishing and subtitle guidelines settle on.
The deeper reason the two numbers are not interchangeable is information density. A Chinese character typically represents a morpheme — a unit of meaning — so a line of 20 characters can carry as much content as a line of 30 to 40 English characters. The eye makes fewer fixations over the same idea. If you estimate Chinese reading time with an English words-per-minute figure, you will overestimate it badly: 600 characters read at 200 "words" per minute looks like three minutes, while at 400 characters per minute it is one and a half.
Reading time estimates are approximations in every language. They assume silent reading of continuous prose, no tables, no code and no re-reading. They are useful for pacing a talk or judging whether an article is a three-minute read or a ten-minute one; they are not useful for calculating a precise delivery time.
The conventions, side by side
| Language | Unit tools count | Reading benchmark | Where the unit shows up |
|---|---|---|---|
| English | Words (space-delimited) | 200–250 wpm | Essay limits, translation quotes, blog briefs |
| Chinese | Characters (字) | ~400 characters/min | WeChat articles, translation quotes, school essays |
| Japanese | Characters (字) | ~400–600 characters/min | Manuscript pages, subtitle timing |
| Korean | Eojeol (space-delimited) | ~300–400 eojeol/min | Publishing and academic limits |
Where the unit becomes money
Translation is priced in the unit the market uses. English-language agencies quote per source word; Chinese, Japanese and Korean agencies quote per character, because characters are what they can count without a segmenter and what their CAT (computer-assisted translation) tools store in translation memory. If you send a 900-character Chinese page to an agency that quotes per word, the number you receive will depend on their internal character-to-word ratio, and that ratio is rarely stated.
Rules of thumb circulate for planning: English words multiplied by roughly 1.5 gives an approximate Chinese character count for the same content, and English text typically expands by 20–35% when translated into German or French and contracts when translated into Chinese. These are planning estimates for layout, not billing figures, and they vary by subject matter — a legal text with long noun phrases behaves differently from a product page.
The same logic explains why platform limits read as they do. A meta description is limited by pixels rather than characters, but the practical budget is around 155 characters of Latin text; the same field in Chinese fits more meaning into the same width. A 280-character social post counts CJK characters as two units on some platforms, so a Chinese post that looks short can be rejected. Always check the platform's own counting rule before trimming, because the number shown in your editor is not necessarily the number the API enforces.
Why two counters give you different numbers
Paste the same paragraph into two word processors and the totals will often differ by a few percent. The usual causes:
- Contractions and hyphenated compounds. Some tools count
don'tas one word, others as two;well-knownis one or two depending on the hyphen rule. - Numbers and units.
1,250may count as one word,1,250.75as one or two, and3 kgas two. - CJK punctuation. Whether 。,、 count as characters changes a Chinese total by 10–15% on ordinary prose.
- Hidden text. Comments, footnotes, headers, text boxes and tracked deletions may or may not be included in a "word count" dialog.
- URLs and code. A pasted URL is one token to a counter and several to a human reader; a snippet of code may contain no "words" at all by one definition.
None of these is a bug. Each tool documents its rule somewhere in a manual nobody reads. The practical response is to standardise on one counter for a project and to state the unit when you hand a number to someone else — "1,840 words" and "1,840 characters" are different deliverables.
How mixed-language text gets counted
A bilingual document has no single correct unit, so the honest approach is to report the parts. TextBench counts Latin-script tokens and CJK characters separately, then adds them for a combined "words" figure and uses both rates for reading time: English words at 200 per minute, CJK characters at 400 per minute, summed. A paragraph of 100 English words plus 200 Chinese characters therefore reads as 0.5 + 0.5 = 1 minute.
If you mostly need one number for a deadline, decide which convention the recipient uses and count in that unit alone. If you need a defensible number for a client, give both counts and the rule you applied. Uncertainty about units causes more disputes than any counting error.
A checklist for bilingual word counts
- Ask which unit the limit or quote uses: words or characters. The answer changes the target by a factor of two or more.
- Strip markup, comments and tracking marks before counting, then count the same way twice.
- For translation estimates, count the source in its native unit and convert with a stated ratio, not a guess.
- For subtitles, work in characters per second rather than words: 15–20 CPS for adult audiences, lower for children's content.
- Record the counter and the unit in your notes, so the next person can reproduce the number.
FAQ
Is one Chinese character equal to one English word?
Not exactly, but the two are close enough that character-based and word-based targets are often treated as comparable. In localization planning, English word counts are usually multiplied by about 1.5 to estimate the Chinese character count of the same content, which implies roughly two-thirds of an English word per character. For billing and for academic limits, always use the unit the other party specified rather than converting.
Do punctuation marks and spaces count as characters?
It depends on the field. Character limits on forms and meta descriptions count punctuation and spaces, because the database stores them. Word processors usually exclude spaces from "characters (no spaces)" and include punctuation in the full count. In Chinese publishing, punctuation is normally counted as characters — a 400-character page includes its commas and full stops.
Why does my editor show a different reading time?
Reading time is a formula, not a measurement, and the formula differs: some tools use 200 words per minute, some 238, some adjust for code blocks or images. If the number matters — for a script read, a podcast segment or a timed exam — calculate it yourself from a stated speed rather than trusting a badge in the corner.
Next: cleaning a pasted list without losing rows, or choosing between the seven case styles. Count your own text with the TextBench word and character counter, which reports characters with and without spaces, words, lines, paragraphs, sentences and reading time at both rates.
Written by the TextBench team. Last updated November 19, 2026.