Counting Chinese · 8 min read
How to Count Chinese Characters: Why Tools Disagree
Word processors, translation agencies and HSK examiners each count Chinese differently. Here is what every number measures and which one to quote.
Published
Paste the same Chinese paragraph into Microsoft Word, into Google Docs, and into a translation agency's quoting form, and you will get three different numbers. None of them is wrong. They are measuring three different things, and the gap between them is large enough to matter when money or a pass mark is involved.
This piece explains what each count measures, where the discrepancies come from, and how to work out which figure the person asking actually wants.
Try it yourself
Count characters, words and sentences in Chinese text.
Open the Chinese Word Counter →Why Chinese resists counting
English writing hands you word boundaries for free. Spaces separate words, so counting words is a matter of counting gaps and adding one. Character counts are equally trivial. The two numbers sit in a stable relationship — an average English word runs about five letters plus a space — so either one tells you roughly what you need to know about length.
Chinese provides none of this. 我今天下午要去图书馆还书 is thirteen characters with no internal punctuation and no spaces, and a reader parses it into words using knowledge rather than typography. Depending on how you segment it, that string contains somewhere between six and nine words. There is no single correct answer, because word segmentation in Chinese is a linguistic judgement, not a mechanical one.
The practical consequence: a Chinese character count is objective and reproducible, while a Chinese word count is an estimate produced by whatever segmentation algorithm your software happens to use. This is precisely why the industries that depend on accurate length measurement — translation, publishing, examinations — standardised on characters and largely ignore words.
The four numbers you will encounter
Almost every dispute about Chinese text length comes down to confusion between these four measures.
- Chinese characters only — Han ideographs, excluding punctuation, Latin letters, digits and spaces. This is the translation industry's standard unit and the one that behaves most predictably.
- Total characters including punctuation — every non-whitespace character. Usually 10-15% higher than the Chinese-only figure, because Chinese punctuation is frequent and full-width.
- Total characters including spaces — adds whitespace and line breaks. In pure Chinese text this barely differs from the previous figure; in mixed Chinese-English text it can be substantially higher.
- Words — the output of a segmentation algorithm. Typically 55-65% of the Chinese character count, but varies by several percentage points depending on the algorithm.
| Measure | Example: 你好，我叫 Mary，今年 25 岁。 | Count |
|---|---|---|
| Chinese characters | 你好我叫今年岁 | 7 |
| Total characters (no spaces) | includes ，。and Mary and 25 | 20 |
| Total characters (with spaces) | adds the two spaces | 22 |
| Words | 你好 / 我 / 叫 / Mary / 今年 / 25 / 岁 | 7 |
What common software actually counts
Microsoft Word's status bar shows a word count that, for Chinese text, is really a character count with Latin words counted as single units. Word's own documentation describes this as counting 'Asian characters' separately, and the detailed word count dialog exposes the breakdown — but the number on the status bar is the conflated one, which is where most confusion starts.
Google Docs takes a similar approach but segments differently, so the two applications disagree on the same file. Neither is authoritative.
Translation memory tools such as Trados and memoQ count source characters, and are generally the closest thing to an industry reference. If a client and a vendor disagree, the CAT tool's analysis report is usually what settles it.
Web forms and CMS fields almost always count JavaScript string length, which is UTF-16 code units. For the overwhelming majority of Chinese characters this equals the character count, but characters outside the Basic Multilingual Plane — rare surname characters, some historical forms, and every emoji — count as two. This is why a name occasionally fails a length check that it visibly passes.
How translation pricing uses the count
Chinese-to-English translation is priced per source Chinese character, typically quoted per thousand. The convention exists because the character count is the only figure both parties can reproduce independently, and because it is stable — the same document counted twice gives the same number.
A useful planning ratio: 1,000 Chinese characters produces roughly 600-700 English words. Technical and legal material lands at the lower end because Chinese compresses terminology efficiently; conversational and literary material lands higher because English needs more function words to carry the same content.
When you are being quoted, confirm which count is being used. A quote based on total characters including punctuation is roughly 12% higher than one based on Chinese characters alone, and on a large project that gap is real money. Ask explicitly: 'is that Chinese characters only, or all characters?'
Exams, subtitles and hard limits
HSK writing tasks specify character minimums — 80 characters at HSK 4, 400 at HSK 5, 600 at HSK 6 — and the examiners count Chinese characters. Punctuation does not count toward the minimum, which matters if you are padding a short essay with short sentences.
Subtitling works to a different constraint entirely: characters per line and reading speed. Chinese subtitles conventionally run to about 14-16 characters per line, and the limit is driven by how fast a viewer can read, not by how much fits on screen.
Social platforms count in their own way. Weibo's 140-character limit counts Chinese characters as one and Latin letters as half, which is why an English post fits far more text than the number suggests.
Getting the number you actually need
Start from the question rather than the tool. If you are being paid, count Chinese characters and confirm that is what the other side is counting. If you are filling a field with a hard limit, count total characters including spaces, because that is what the validator will do. If you are estimating how long something takes to read or say, character count is the right basis — words are too unstable to plan against.
The one case where word count genuinely helps is comparing Chinese text against an English original or translation, since English length is naturally measured in words. Even there, treat the Chinese word count as approximate and use the character count as the contractual figure.
Key takeaways
- Chinese character counts are objective; Chinese word counts are algorithm-dependent estimates.
- Translation is priced per source Chinese character, usually per thousand.
- 1,000 Chinese characters translates to roughly 600-700 English words.
- HSK writing minimums count Chinese characters, not punctuation.
- Web form limits count UTF-16 units, so rare characters and emoji count double.
Frequently asked questions
Does punctuation count as a character?
It depends who is asking. Translation quotes and HSK writing minimums exclude it; web form validators and 'total character' counts include it. Chinese punctuation is full-width and frequent, so the difference is typically 10-15%.
Why do Word and Google Docs disagree on the same document?
They use different segmentation rules for mixed Chinese-Latin text and treat punctuation differently. Neither is authoritative; use a tool that shows you the breakdown so you can see which component is causing the gap.
How many characters is a page of Chinese?
A typical A4 page of body text holds roughly 700-900 Chinese characters at normal margins and 12pt type. Academic formatting requirements in mainland China often specify characters per page directly.