How to learn Chinese from social media

Start with comments, not articles: they are short and use the words HSK teaches first, plus a few dozen spoken words HSK leaves out.

By Mandarin The WildUpdated 20 October 20265 min read

We make Mandarin The Wild.

Where Chinese is written every day

At the end of 2025, 1.125 billion people in China used the internet, and 1.074 billion of them watched short videos, 95% of all users.¹ Every video, song and post gathers comments, and on Bilibili the comments run across the video itself as danmu (弹幕). That is the largest body of everyday written Chinese there is, and most of it is short.

我看视频的时候喜欢开着弹幕。

dànmù: “I like having the danmu on when I watch videos.”

弹dàn幕mù

Noun

bullet comments; danmaku (comments scrolling over a video)

Also written 彈幕

What the studies found

People learn words just from meeting them in what they read, watch and listen to. Across 24 studies of this incidental learning, learners knew 9–18% of the new words on a test right after, and they kept more of them when they met the words over several sessions rather than in one. Viewing alone gave the smallest gains.² Learning words with technology, from apps and games to messaging, had a moderate effect across 34 studies with 2,511 learners.³

Chinese online writing has its own habits. In 331 examples of internet slang on Douyin, the most common devices were English-Chinese mixes, pinyin initials (such as yyds) and words that mix characters, letters and numbers, followed by puns on how words sound and spoken particles.⁴ That is the part a textbook won't prepare you for, so we measured how big it is.

How much of it can a learner read?

We read 231,947 words of posts from five places, cut them into words and checked each word against the 2025 HSK list, leaving out names.⁶⁷.⁸.¹⁰.

Where	Posts	Characters a post (median)	HSK 1–3	HSK 1–4	HSK 1–6
Weibo posts	4,298	13	63%	74%	87%
Bilibili comments	6,946	14	65%	75%	88%
Bilibili and AcFun danmu	8,682	9	66%	75%	87%
NetEase Cloud Music comments	3,169	13	68%	79%	90%
YouTube comments in Chinese	642	16	65%	76%	90%
All five	23,737		65%	75%	88%
Web text, counted the same way			47%	60%	79%

Social media is far easier than the web around it. HSK 1–4 words cover 75% of it, nearly as much as of a TV show's subtitles (79%) and far more than of the news (52%) or a phone's menus (56%). Song comments on NetEase are the easiest, Weibo posts the hardest. A post is short, so a word you don't know is one of a dozen, and you can look it up and move on.

Even all of HSK 1–6 leaves about one word in eight unknown, and comfortable reading takes 95–98%.⁵ The gap is smaller than it looks, because it is made of a small set of words.

The words HSK leaves out

The commonest words in these posts that HSK 1–6 doesn't teach, each used under at least 20 different videos or threads, are spoken words and the vocabulary of the platforms themselves:

Spoken question words and particles: 啥 (what), 咋 (how), 为啥 (why), 么, 呗, 咯 and 诶. 啥 alone came up 201 times under 187 videos and threads.
Talk about people: 哥, 姐, 大哥, 哥们 (buddy), 俺 and 咱 (I, we).
The platforms' own words: 弹幕 (danmu), 博主 (blogger), 粉丝 (fans), 剧情 (plot) and 微博 (Weibo).
Reactions: 爽 (feels great), 真心 (sincerely), 玩意 (thing, often rude) and 绷不住 (can't keep a straight face).

Fewer posts use letters than you might expect: 4–7% of them contained a word written in Latin letters. The commonest were up (as in UP主, a video's creator), ai, bgm (background music), doge, cp (a couple fans pair up), byd and nb. Our slang page explains the common ones.

A plan that fits the evidence

Get to about HSK 3 first. Even at HSK 3, about a third of the words are new; below it, more. Graded stories in the Library get you there faster, and the level test shows where you are.
Read the comments under things you already watch or listen to. They are short, they repeat each other and you know what they are about.
Learn the spoken words above early. Each of them came up under at least 20 different videos or threads.
Come back day after day. Learners keep more of the words they meet over several sessions than in one long one.
Look words up in place. On a computer, Mandarin The Wild explains the Chinese on a web page under your pointer; on a phone, paste a post into the reader. Save the words that keep coming back to your Backpack and review them.

How we counted

The posts are the simplified Chinese training rows of our text supply, which we collect to test our reader: Weibo posts from a public research set, and Bilibili comments, Bilibili and AcFun danmu, NetEase Cloud Music song comments and Chinese-language YouTube comments collected in October 2026. Words are cut by our engine and levelled on the 2025 HSK list the way the Library levels a story; a run of one repeated character (哈哈哈哈) counts as one word, and names the dictionary only knows as names are left out. Web text is the Mandarin part of FineWeb-2, counted the same way.⁹ A word in the list above had to appear under at least 20 different videos or threads, so one video's running joke doesn't count.

Questions

Can you learn Chinese from social media?

Yes, once you know most of the words. In our count of 23,737 Chinese posts and comments, HSK 1–4 words are 75% of the words, against 60% of web text, and a typical post is 9 to 16 characters long.

What Chinese social media should I use to learn?

Start with the comments under videos and songs you already like, on Bilibili or NetEase Cloud Music: they were the easiest in our count (75–79% HSK 1–4). Weibo posts were a little harder.

What Chinese internet words does HSK not teach?

Mostly spoken words: 啥 (what), 咋 (how), 为啥 (why), 呗, 哥们 (buddy), 咱 (we), and platform words such as 弹幕 (danmu), 博主 (blogger) and 粉丝 (fans).

Sources

Checked on 20 October 2026. Prices and features change; tell us if one has.