Chat moderation has always been issue an for online streaming. It shouldn’t surprise you that it’s also an issue in China, where internet censorship is everywhere. And now it might come to the country’s unique video comment system: Bullet chat.

A new article by China’s communist mouthpiece, the People’s Daily, is urging video platforms to moderate bullet chat, the most popular form of video commentary in the country.

What is it? Imagine the live comments you see on Facebook Live or Twitch -- but instead of being in a little window at the side, the comments are displayed directly on the video itself. They’re pegged to specific moments, and float directly above the footage.

Sometimes people end up enjoying the chat more so than the video. (Picture: Bilibili)

Bullet chat is hugely popular in China. Video site Bilibili, where millions of people watch video with bullet chat every day, became became a Nasdaq-listed company in less than 9 years.

In the recent article, the People’s Daily doubled down on its request for video platforms to better moderate chat. “It is beneficial to create better quality bullet chat if platforms review comments first before airing them,” says the story, which also re-stresses that video platforms should manage bullet chat in real time.

But it’s worth pointing out that bullet chat isn’t completely uncensored today.

How? According to one of the top streaming sites, Douyu, platforms usually use chat filters, hire people to moderate and, lastly, act on users’ complaints.

Chat filters can block sensitive or inappropriate words and meaningless special characters. They can also automatically blacklist user accounts and their IPs.

Now sometimes, the inappropriate speech isn’t in the form of a single word, but a phrase or sentence. How do they ban that? Douyu said its filter can recognize patterns of speech. For example, if the site wants to block a violence-inducing phrase like “Fire first, ask permission later,” it will ban any statement that has “fire” and “permission later”. It doesn’t matter how you switch up the expression or add filler words in between, the system can detect that you are trying to say the banned phrase.

These are all just default filters executed by machines. If these filters fail, Douyu said there is also manual moderation, executed by real human beings.

And if all those still fail? Douyu said it also has a reporting system in place for users to flag inappropriate speech.

Some of these methods are by no means unique to China. Twitch also has a profanity filter and allows streamers to ban undesired comments. But there’s also a difference between Twitch and Chinese streaming platforms.

For one, politically sensitive terms are blocked in China, but not in the West. And Twitch recently shut down its moderation team, allowing channels to appoint their own. Before that, the company had 16 Global Moderators. In China? Bytedance, the company behind viral short video app TikTok, now has 10,000.

This is just part of a larger push to to clean up content online, introduced by Chinese authorities in January. They also require video content to be fully reviewed before it’s broadcast.