An analysis shows how Chinese-language bots on Twitter drowned out tweets about protests over China's zero-COVID-19 policy, as people sought to evade censors
Twitter and its new owner, Elon Musk, have recently vowed to crack down on bots. But the flood of spam for Chinese users …
New York Times
Context & Ripple Effects
The bot flood is the escalation of a fight already documented in the corpus: in early December, Twitter users in China were tunneling past the Great Firewall and manually fighting spam to push protest footage out of the country. The new analysis shows that fight being lost at scale — spam volume, not individual takedowns, became the suppression mechanism.
The pattern has deep roots on both sides of the firewall. Twitter removed [[a:954591|more than 170K accounts tied to the Chinese government pushing deceptive COVID and Hong Kong narratives]] in 2020, while Citizen Lab found over 2,000 coronavirus-related terms censored on WeChat the same year. What changed by late 2022 is the target: not domestic feeds but the offshore platform where censored speech escapes.
First-order effects
Chinese speakers seeking to share zero-COVID protest material on Twitter lose their distribution channel in real time — their tweets are buried under bot spam rather than deleted, making the suppression harder to attribute and contest.
Elon Musk's public vow to crack down on bots faces its sharpest test yet: the failure is visible precisely where Twitter's value as an uncensored channel for Chinese dissent is highest.
Second-order effects
Suppression by flooding compounds quieter enforcement: weeks after this analysis, [[a:835794|over 30 prominent Chinese dissidents reported disappearing from search or unexplained suspensions]], meaning activists were hit by both algorithmic invisibility and spam drowning simultaneously.
If flooding works where deletion fails, state-aligned influence operations shift from account networks that can be purged (as the 170K-account takedown showed) toward volume-based tactics that survive moderation — raising the cost of distinguishing authentic speech from synthetic noise for every platform hosting dissident communities.
Censorship becomes extraterritorial: rather than only policing domestic platforms like WeChat, the playbook extends to shaping what foreign audiences see on Western platforms, turning those platforms' moderation capacity into a geopolitical variable.
The trend: State-aligned information control is moving from deleting banned speech to drowning it in synthetic volume, making platform bot defenses a frontline of cross-border censorship.
New from @elegant_wallaby and @ttzhang107: a look into the wave of spam that drowned out Chinese-language content at the height of the recent anti-COVID lockdown protests in China. https://cyber.fsi.stanford.edu/ ...
The findings match a report published today by David Thiel from @stanfordio who reviewed millions of tweets by searching for 30 Chinese cities and found that bots were active before the protests began and continued after they had ebbed. https://cyber.fsi.stanford.edu/ ...
The @nytimes did their own analysis, cross-checking with SIO. We both conclude that it's a real problem, but probably not linked to the Chinese government or to the protests themselves. https://www.nytimes.com/...
Twitter users were drowned with spam when they searched for information about the historic anti-lockdown protests in China. Through data analysis and interviews with people behind bots, we found that much of the spam is linked to commercial bot networks. https://www.nytimes.com/.…
“In retrospective research, historical Twitter data generally becomes ‘cleaner’ — some amount of spam and inauthentic behavior will have been removed ...ToS-violating or inauthentic content tends to appear most prevalent in the immediate past.” https://cyber.fsi.stanford.edu/ ...
Governments can do brute-force social media network manipulation. They have the infrastructure and the staff to support it. Our best hope is: The platforms stop it, or the media expose it. But not after the fact. https://twitter.com/...
Important read. The spam networks are still active - go search 上海 - and content shows up in Top even absent engagement. Enforcement takedowns seem slower. Attributing something to a state is tough; outside researchers usually collaborate with platform integrity teams on that. htt…
Our team looked at recent spam on Twitter in Chinese and determined that it is probably more a reflection of problems at Twitter fighting spam than a government operation. Also, good lessons on the risks of looking backwards at any abuse on platforms. https://twitter.com/...
Remember that two missions of this site include stopping bots and protecting freedom of expression. These bots went unchecked and drowned out freedom of expression. https://www.nytimes.com/...