Chinese Accounts Use AI to Generate Low-Quality Content and Manipulate Platform Algorithms
A Chinese blogger disclosed in March 2026 that thousands of low-quality accounts coordinated across X platform to systematically suppress their dissenting content by exploiting recommendation algorithms. China's government simultaneously announced removal of 39,000 accounts and 708,000 violations for using AI to batch-generate low-quality content and misinformation.
What happened
In March 2026, a Chinese blogger on X platform publicly disclosed experiencing algorithm manipulation where thousands of low-quality accounts coordinated to unfollow and block the account in stages. The blogger described a 'three-stage attack' model: accounts without avatars or posts suddenly followed the target account ('embedding'), remained dormant to link the account to a low-quality social graph ('lurking'), and then collectively unfollowed and blocked when political content was trending ('detonation').
China's Central Network Information Office announced on March 10, 2026 that it had disposed of 39,000 accounts and removed 708,000 violations during a 'Clear and Bright 2026' campaign. Accounts including 'Fat Grandpa AI' on Kuaishou and 'Couple Bed Collapse Diary' on Douyin were using AI to batch generate low-quality, incoherent content such as 'Fat woman fell in water' and 'Big grandpa fell in pit'. Other accounts used AI to modify classic literature with graphic, grotesque imagery labeled as children's animation.
Reported impact
The targeted blogger experienced a dramatic drop in post reach after the coordinated unfollows and blocks triggered the platform's algorithm to suppress content distribution. Accounts placed on search blacklists became difficult to find even when directly searched by ID. The blogger reported that traffic to subsequent posts experienced a sharp decline, creating an 'invisible censorship' effect while appearing to result from platform rules rather than human action.
Multiple commentators including media personality Wenzhao noted that large-scale traffic manipulation has been occurring for over two years, causing content creators on X and YouTube to experience abnormal traffic patterns. The widespread dissemination of AI-generated low-quality content and graphic modified literature contaminated the platform ecosystem and disrupted normal user experience.
Index analysis
The attack exploits X platform's recommendation algorithm logic by creating the appearance that content was flagged as 'extremely offensive' or containing false information when large numbers of followers leave during peak engagement. Bulk blocks signal account violations in the algorithm's reputation weighting system, potentially triggering search blacklisting mechanisms that platform systems cannot distinguish from genuine user behavior.
The coordinated timing of actions concentrated during Beijing business hours, combined with the use of low-quality accounts, suggests systematic coordination. Observers noted that with the proliferation of AI tools, the threshold for large-scale manipulation has significantly lowered in terms of cost and technical difficulty.
Governance controls
The targeted blogger recommended manually removing 'zero-content accounts' without avatars, posts, or recent registration dates to reduce algorithmic association with low-quality networks. When facing concentrated attacks, the blogger suggested increasing posting volume and engaging with real followers' positive interactions to offset negative signals.
The blogger and other observers emphasized that platform design rules are based on normal civil order and lack effective responses to systematic non-conventional manipulation. Commentators suggested maintaining content quality and fan loyalty as the primary defense mechanism against coordinated operations.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.