How to read Chinese: from pinyin to your first real book
To read Chinese without help you need about 98% of the words on the page. We measured what each HSK level gets you in stories, TV and the news.
Learn the words for HSK 4 and you'll know about 84% of the words in Chinese films and TV shows. In a news report, you'll know 57%.
This guide is for you if you've started Chinese and want to know what you can read now, what to read next, and how to tell when a book is within reach.
Key takeaways
- Reading means knowing words, not just characters. You can know sky, 天, and air, 气, and still not know weather, 天气.
- You need nearly every word to read on your own. Below about 98% known words, understanding falls away steadily.
- Graded stories are the bridge. For years they're the only texts most learners can read, so the help they give matters.
- TV dialogue is the easiest native text we measured. It passes 95% known words with the whole HSK list, and the news never does.
- Test a page before you start a book. Paste it into the checker on this page and see how much of it you'd know.
Reading Chinese means knowing words
When you read Chinese, you read words, not characters one at a time. About three in four words on the HSK lists are two characters long, by our count.
A word can mean something its characters don't spell out. Thing, 东西, is written with east and west. You might know both characters in weather, 天气, and still stall on it the first time.
That's why the research on reading counts words, and so do we. Characters matter too, and our guide to how many Chinese characters you need counts those. Here, the question is simpler: of the words on this page, how many do you already know?
All our counts use Simplified characters, the script of the HSK. If you'll read Taiwan or Hong Kong texts, our guide to Simplified vs Traditional Chinese helps you choose.
You need about 98% of the words
To read a text without help, you need to know about 98% of its words: one new word in 50.
Drop to 90% and it's one in 10. In Chinese, that means several new words in every paragraph.
The classic test comes from Marcella Hu and Paul Nation. They gave 66 learners the same short story, with different shares of its words swapped for made-up ones. Here's how many understood it well:
Source: Hu & Nation (2000), Table 4. The readers were adults on a pre-university English course, reading a 673-word story.
Their conclusion: most readers need about 98%. A later study put the minimum at 95% and kept 98% as the ideal.
Since then, the line has softened but the lesson hasn't. A study of 661 learners in eight countries found understanding rising steadily with every extra word known, with no sudden cliff. A 2023 replication of Hu and Nation's study couldn't fully reproduce its results.
It holds for beginners too. In a study of first-year Spanish learners reading graded texts, knowing 98% of the words went with remembering most of the story. An eye-tracking study found signs that reading at 98% takes less effort than at 90% or 95%.
One caveat: nearly all this research is on learners of English. We found no experiment like it with learners of Chinese. A 2025 study of Chinese textbooks did use the same yardstick, and found that by the final volume, the words they teach cover 81% of real texts.
So treat 95% as the floor for reading on your own. At 98%, reading starts to feel easy.
What can you read at each HSK level?
Graded stories are the only texts most learners can read for their first few years. Native text is much harder, and the news is the hardest we measured.
We counted, for each HSK level, the share of words a learner would know in five kinds of text. A word counts as known if it's on the HSK list at or below that level, or built from words that are. Names are left out. Here's what the HSK 4 list gets you, and what the whole list gets you:
Only graded stories and TV dialogue pass 95%
Share of running words a learner would know, names left out
- 95–98%: reading on your own
| Text | At HSK 4 (2,000 words) | With every HSK word (11,000) |
|---|---|---|
| Graded stories (SayMei) | 91.6% | 98.4% |
| TV and film dialogue | 84.0% | 97.0% |
| News (Wikinews) | 56.6% | 93.8% |
| Encyclopedia (Wikipedia) | 59.5% | 92.6% |
| Classic fiction (Lu Xun, 1923) | 70.5% | 91.9% |
Source: our count over SayMei's 166 graded books, 6,243 film and TV subtitle files, 8,789 Wikinews articles, 4,997 Wikipedia sentences and Lu Xun's Call to Arms, using the HSK 3.0 word lists (2025). Method at the end.
Even someone who knows every HSK word can't read the typical news article on their own.
Only about one news article in three reaches 95% for that reader. Names, places and specialist terms make up the gap.
Here's how the climb looks level by level, for the three kinds of text most learners want to read:
| You know | Graded stories | TV and film | News |
|---|---|---|---|
| HSK 1, 300 words | 66.0% | 57.9% | 25.3% |
| HSK 2, 500 words | 78.5% | 67.8% | 30.6% |
| HSK 3, 1,000 words | 88.2% | 76.3% | 39.7% |
| HSK 4, 2,000 words | 91.6% | 84.0% | 56.6% |
| HSK 5, 3,600 words | 94.4% | 89.3% | 73.3% |
| HSK 6, 5,400 words | 96.7% | 92.8% | 82.5% |
| HSK 7–9, 11,000 words | 98.4% | 97.0% | 93.8% |
Source: our count. Graded stories: the median of SayMei's 166 books. TV and film: all 33.5 million words of the SUBTLEX-CH subtitles. News: the median of 8,789 Wikinews articles.
If you're at HSK 3, you know nearly nine words in ten of a typical graded story. That's why pinyin, the English line and the audio aren't crutches at that level. They're how reading works.
Check any text against your level
The checker does the same count for any text you paste in, using your browser's own word splitter. Try a page from the book you want to read:
How readable is this for me?
Paste any Chinese text, pick your level, and see the share of its words you'd know. At HSK 4, that's 95.5% of this graded story, 71.8% of Lu Xun and 61.1% of a news report.
Text
61.1%of the words known at HSK 4
35 of 90 words are new at HSK 4: about 1 in every 3.
Study, not reading. With more than one new word in 10, you'll spend your time looking things up. Pick something easier, or learn the repeated words first.
Go one rung down the ladder: a graded story written for your level. Find a story at your level
This text at every HSK level
Stephen Hawking visits Hong Kong 霍金访问香港
英国著名科学家斯蒂芬·霍金12日下午抵达香港，开始为期6天的访港之旅。霍金将在香港科技大学主持一场以“宇宙的起源”为题的公开讲座讲述有关宇宙起源的理论。坐着电动轮椅的霍金在机场受到了热烈欢迎。参加欢迎仪式的小学生们手持横幅，齐声用英语高呼“欢迎霍金”，霍金则通过电脑语音合成器说：“非常感谢”。
New at your level Beyond the HSK lists (often a name)
New words in this text, most repeated first
- 霍
- 港
- 起源
- 宇宙
- 讲述
Names count as new words until you tap them, so news reports come out a little low. The text never leaves your device.
Climb the reading ladder one rung at a time
You learn to read in five rungs, and each one takes thousands more words than the last. Here's how many words each kind of text takes, from readable with help to readable on your own:
Each rung of the ladder takes thousands more words
Words you'd need, learning the HSK words in order, for the typical text to reach 90% (bar start) and 95% (bar end)
| Kind of text | Words you need | Where that falls |
|---|---|---|
| Graded stories | 1,500–4,000 | With help from HSK 4; on your own between HSK 5 and 6 |
| TV and film dialogue | 4,000–8,500 | With help between HSK 5 and 6; on your own within HSK 7–9 |
| News | 9,000–11,000+ | With help near the end of HSK 7–9; on your own, beyond the lists |
| Classic fiction | 9,500–11,000+ | With help near the end of HSK 7–9; on your own, beyond the lists |
Source: our count, as in the chart above. Estimates between the HSK levels we measured, rounded to the nearest 500 words.
The biggest gap sits between graded stories and everything native, so it's worth crossing in small steps. Here are the rungs, and what helps on each:
- The sounds. Learn pinyin, the official spelling of Chinese sounds since 1958, and the four tones. Aim to read any syllable aloud. Our free pinyin chart has audio for every one.
- Graded stories, with help. At HSK 1 to 3, read with pinyin over the characters, an English line and the audio. In a 26-week study, beginners who read with audio gained more speed and understanding than silent readers.
- Graded stories, on your own. Around HSK 4 and 5, turn pinyin off or show it only for new words. An experiment found partial pinyin worked best once learners had some experience.
- Your first native texts. Around HSK 6, start with dialogue: dramas with Chinese subtitles, comics, and web novels in a reader with a tap dictionary.
- News and literature. Expect to look things up even after the HSK lists end.
Don't skip rungs. Reading at 80% known words isn't reading. It's decoding, and it's slow enough to make anyone quit.
How do you choose your first real book?
Choose it by coverage, not by title. Paste a page into the checker and look for 95% or better at your level. At 98%, you'll read for pleasure.
Some famous books are harder than they look. Lu Xun is one of China's best-known modern writers, and his stories are short. Even with every HSK word, you'd know only about 92% of their words. Here's how his story "My Old Home" begins:
我冒了严寒，回到相隔二千余里，别了二十余年的故乡去。
Wǒ màole yánhán, huídào xiānggé èrqiān yú lǐ, biéle èrshí yú nián de gùxiāng qù.
"Braving the bitter cold, I went back to my old home, more than two thousand li away, which I had left more than twenty years before."
What makes a first book easier:
- Dialogue and everyday life. In our count, film and TV dialogue beat the news and the encyclopedia at every level.
- Modern Chinese. Writing from the 1920s uses words and forms today's lists don't teach.
- A story you already know. In Hu and Nation's study, the few readers who coped with unknown words leaned on reading skill and background knowledge.
- A reader app. A tap dictionary on native web novels makes the jump smaller; our story and news app comparison covers the apps.
If a page comes out under 90%, park the book. Read graded stories for a level or two, and try it again.
SayMei Stories covers the middle rungs
Our graded stories are built for rungs two and three. SayMei Stories has 166 graded books in 812 short chapters, levelled from about HSK 1 to HSK 4. Every line has pinyin, an English translation and audio, and you can tap any word for its meaning.
The first two chapters of every book, and 8 whole books, are free with no sign-up. The full library costs $2.99 a month after a free week, which needs a card.
Where it falls short:
- Its levels are generous. A book's own-level words cover well under 95% of it, so read with the help switched on. Our stories guide has the numbers.
- It stops around HSK 4, and it has no native texts.
- It's one library among many. The story and news app comparison covers the others.
Free tools help with every rung: the pinyin chart, the HSK word lists, the level test and a Chinese to pinyin converter for any text.
What to do next
- Find your level. Take the free level test, so you know which word list you're on.
- Read a graded story every day. Keep pinyin and audio on until a page feels easy.
- Test before you commit. Paste a page of any book or article into the checker above.
- Learn the words that repeat. Move the checker's new words into flashcards.
- Climb when it's easy. When you rarely need to tap a word, move up a level.
Start with chapter 1 of any SayMei story: it's free, with pinyin, translation and audio on every line.
FAQ
How long does it take to learn to read Chinese?
It depends on what you want to read. Graded stories are readable with help from your first months. By our count, film and TV dialogue reaches 95% known words at about 8,500 words, near the end of the HSK lists, and the typical news article doesn't get there even then. Our guide to how long Mandarin takes turns levels into hours.
Can I read Chinese after HSK 4?
Graded stories, yes. Native texts, only with help. With HSK 4's words, you'd know about 92% of the words in our graded stories but only 57% in a typical news report, by our count. Reading on your own takes nearly every word, so use pinyin, audio and a tap dictionary for anything native at that level.
At what HSK level can you read a novel?
With a tap dictionary, around HSK 6 is a realistic start; on your own, later. We couldn't count modern novels, which are under copyright. But at HSK 6 you'd know about 93% of the words in film and TV dialogue, and even with every HSK word only about 92% of Lu Xun's classic stories. Paste a page of your novel into the checker to see where it falls.
When should I switch from graded readers to native material?
When a native text gives you about 95% known words, the floor for reading on your own. For film and TV dialogue, that comes between HSK 6 and the end of the HSK lists, by our count. Until then, mix the two: graded readers for volume, and short native pieces with a dictionary as a stretch.
Should I read Chinese with pinyin?
Yes at first, then less. Pinyin gets beginners reading real sentences, but beginners tend to read it instead of the characters: in an eye-tracking study, all ten learners looked at the pinyin when it was there. For learners with some experience, pinyin over new words only worked best. Try reading each line twice, the second time without it.
How many words do you need to read a Chinese newspaper?
More than the HSK lists hold. Even with every HSK word, the typical Wikinews article stays below the 95% line, by our count, and only about one article in three crosses it. Names, places and specialist terms make up the rest, so even advanced readers keep a dictionary close. For characters rather than words, see our characters guide.
Is reading Chinese harder than speaking it?
For most English speakers, yes, at least at first. In a study of US college students, learners of Chinese spoke about as well as learners of Spanish after two years. But they read at a novice level, while the Spanish learners read at an intermediate one. Graded stories with audio help, because you hear each word as you read it.
How can I read Chinese faster?
Read more easy text, and read along with audio. In a 26-week study, beginners who read graded readers with the audio improved their reading speed and comprehension more than silent readers, and kept the gains months later. Choose texts at about 98% known words, where you read for meaning instead of decoding.
Sources
- Hu, M. H.-c., & Nation, P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1), 403–430. https://doi.org/10.64152/10125/66973
- Schmitt, N., Jiang, X., & Grabe, W. (2011). The percentage of words known in a text and reading comprehension. The Modern Language Journal, 95(1), 26–43. https://doi.org/10.1111/j.1540-4781.2011.01146.x
- Laufer, B., & Ravenhorst-Kalovski, G. C. (2010). Lexical threshold revisited. Reading in a Foreign Language, 22(1), 15–30. https://doi.org/10.64152/10125/66648
- Kremmel, B., Indrarathne, B., Kormos, J., & Suzuki, S. (2023). Unknown vocabulary density and reading comprehension: Replicating Hu and Nation (2000). Language Learning, 73(4), 1127–1163. https://doi.org/10.1111/lang.12622
- Pellicer-Sánchez, A., Webb, S., & Wang, A. (2024). How does lexical coverage affect the processing of L2 texts? Applied Linguistics. https://doi.org/10.1093/applin/amae062
- Herman, E., & Leeser, M. J. (2022). The relationship between lexical coverage and type of reading comprehension in beginning L2 Spanish learners. The Modern Language Journal. https://doi.org/10.1111/modl.12761
- Shen, M., Niu, R., & Liu, H. (2025). Cumulative coverage of Chinese L2 textbooks: How well do they prepare students for reading real-world texts? Language, Culture and Curriculum. https://doi.org/10.1080/07908318.2025.2528779
- Chang, A. C.-S., & Millett, S. (2015). Improving reading rates and comprehension through audio-assisted extensive reading for beginner learners. System, 52, 91–102. https://doi.org/10.1016/j.system.2015.05.003
- Lee, C. H., & Kalyuga, S. (2011). Effectiveness of on-screen pinyin in learning Chinese. Computers in Human Behavior, 27(1), 11–15. https://doi.org/10.1016/j.chb.2010.05.005
- Stickler, U., & Shi, L. (2015). Eye movements of online Chinese learners. CALICO Journal, 32(1), 52–81. https://researchonline.lse.ac.uk/id/eprint/60826
- Winke, P., Zhang, X., Rubio, F., Gass, S., Soneson, D., & Hacking, J. (2020). The proficiency profile of language students. Second Language Research & Practice, 1(1), 25–64. https://scholarspace.manoa.hawaii.edu/items/1ab50fef-1247-494e-8af8-c1a79026dd85
- Cai, Q., & Brysbaert, M. (2010). SUBTLEX-CH: Chinese word and character frequencies based on film subtitles. PLoS ONE, 5(6), e10729. https://doi.org/10.1371/journal.pone.0010729
- HSK 3.0 syllabus (2025), Center for Language Education and Cooperation. https://hsk.cn-bj.ufileos.com/3.0/%E6%96%B0%E7%89%88HSK%E8%80%83%E8%AF%95%E5%A4%A7%E7%BA%B21219.pdf
- Universal Dependencies, Chinese GSDSimp treebank, v2.17 (CC BY-SA 4.0). https://github.com/UniversalDependencies/UD_Chinese-GSDSimp
- Chinese Wikinews, articles dump of 1 October 2026 (CC BY 2.5). https://zh.wikinews.org/
- Lu Xun, 呐喊 (Call to Arms, 1923), on Chinese Wikisource (public domain). https://zh.wikisource.org/wiki/%E5%90%B6%E5%96%8A
- 汉语拼音方案 Hànyǔ pīnyīn fāng'àn (Scheme for the Chinese Phonetic Alphabet), approved 11 February 1958, on Chinese Wikisource. https://zh.wikisource.org/wiki/%E6%B1%89%E8%AF%AD%E6%8B%BC%E9%9F%B3%E6%96%B9%E6%A1%88
- SayMei Stories library and pricing, checked 8 October 2026: /stories, /pricing.
Methods
We counted the share of running words a learner would know at each HSK level, in five collections of text, on 8 October 2026.
Word lists. The HSK 3.0 syllabus (2025), cumulative: 300 words at HSK 1 up to 11,000 at HSK 7–9. A few short forms the syllabus prints only inside longer entries, such as 妈 inside 妈妈, take their level from the 2021 Standard. A word with an optional 儿 also counts without it.
What counts as known. A word on the list at or below the level, or a word that splits into such words, such as 走了, left, which is 走 plus 了. Chinese numerals count as known. Names are left out of the count.
The texts.
- Graded stories: all 166 SayMei books, using the reader's own word boundaries and name markings. The typical book is the median.
- Film and TV dialogue: the SUBTLEX-CH word list of 33.5 million words from 6,243 subtitle files, with names left out by their part-of-speech tags.
- News: the 8,789 Chinese Wikinews articles of 100 characters or more in the 1 October 2026 dump, converted to Simplified characters and split into words with the jieba segmenter. The typical article is the median.
- Encyclopedia: the 4,997 Wikipedia sentences of the Universal Dependencies GSDSimp treebank, split into words by hand.
- Classic fiction: the 14 stories of Lu Xun's Call to Arms from Wikisource, converted to Simplified characters, with 56 old print variants written in their modern forms, split with jieba.
Words you need. For the ladder chart, we estimated the words needed for the typical text to reach 90% and 95% by drawing a straight line between the HSK levels we measured, then rounded to the nearest 500.
The checker splits words with your browser's built-in word splitter and counts them the same way, but it can't tell names from words. On the same stories, news and fiction, its results came within about 3 points of the figures above.
Limits. An HSK list is a stand-in for what a real learner knows. Word boundaries in Chinese are fuzzy, so different splitters give slightly different counts. Counting words built from known words as known is generous: you can often guess them, but not always. No learner data was used.
Every Mandarin example was checked against the CC-CEDICT dictionary. If you spot a mistake, tell us.
