A study published on Thursday (August 8) claims that TikTok has been manipulating its algorithm to promote official Chinese Communist Party (CCP) rhetoric and suppress content related to anti-CCP sentiments. This finding echoes long-standing concerns of the U.S. government that hostile nations are exploiting freedom of speech to undermine national security.
The research report titled “The CCP’s Digital Charm Offensive,” by the Network Contagion Research Institute (NCRI), found that most pro-CCP content originates from entities associated with the Chinese government, including media organizations and influential figures.
For example, some pro-CCP “borderland influencers” (travel video bloggers) spare no effort in posting information about the customs and culture of regions like Xinjiang but conveniently omit the fact that the Chinese government has interned over a million Uighur and other minority ethnic groups in “re-education camps” there.
To conduct this study, analysts created 24 new accounts on the short video applications TikTok, YouTube, and Instagram under ByteDance and collected data through eight “sock puppet accounts” pretending to be 16-year-old American users.
These accounts searched for four typical CCP political “sensitive words” – “Tiananmen Square,” “Tibet,” “Uighurs,” and “Xinjiang” – collecting over 3400 relevant search results. Subsequently, human auditors categorized the content by video into “pro-CCP,” “anti-CCP,” “neutral,” or “irrelevant.”
Analysis of these search results indicated that TikTok displays a much higher proportion of “pro-CCP,” “neutral,” or “irrelevant” content compared to Instagram and YouTube. Conversely, the percentage of “anti-CCP” content is significantly lower on TikTok compared to the other two platforms.
For instance, in search results related to “Tibet,” TikTok showed a 30.1% proportion of “pro-CCP” videos, much higher than YouTube’s 13.7% and Instagram’s 4.7%. In contrast, TikTok only displayed 5% “anti-CCP” videos, much lower than YouTube’s 12% and Instagram’s 31.7%.
The analysis suggests that TikTok is suppressing anti-CCP content to weaken criticism of the Chinese authorities. TikTok contains the lowest proportion of anti-CCP content compared to YouTube and Instagram. Additionally, 61%-93% of TikTok’s search results show pro-CCP or irrelevant content, indicating a systematic amplification of pro-CCP and distracting irrelevant content to overshadow the minimal anti-CCP content on the platform.
The study also found that TikTok engages in “tag hijacking” by attaching popular tags (keywords) of highly sensitive or provocative topics according to the Chinese government on unrelated posts, attempting to deceive and drown out real content.
In search results for “Uighurs” on TikTok, the proportion of videos labeled as “irrelevant” reached 60.3%, while on Instagram and YouTube, this proportion is less than 4%. This indicates that TikTok adopts tag hijacking strategies to reduce the impact of anti-CCP narratives.
Despite the data analysis revealing TikTok’s pro-CCP algorithm bias, analysis of user engagement (likes and views) on content related to the four CCP political “sensitive words” showed that, except for “Tibet,” anti-CCP content received significantly higher user engagement than pro-CCP content.
In searches related to “Xinjiang,” the proportion of pro-CCP content skyrocketed on YouTube and Instagram, reaching 52.6% and 49.3%, respectively, compared to only 24% on TikTok. Researchers attribute this anomaly to the combination of pro-CCP media assets – including borderland influencers and official media accounts – dominating social discussions related to these keywords, thus subverting search results.
On YouTube, four out of the five most active accounts in the search results are CCP official or semi-official media accounts (CGTN, “South China Morning Post,” “Shanghai Eye,” and CCTV+), collectively amassing 8 million followers.
Their dominance in the “Xinjiang” search results on YouTube is evident. These four pro-CCP accounts generated 21.7% of the “Xinjiang” search content on YouTube and nearly 40% of pro-CCP content.
Joel Finkelstein, a senior researcher at Rutgers University and the NCRI’s director and chief scientist, led the research team and stated in the report: “The unusual aspect of TikTok is that accurate information on CCP human rights violations is systematically pushed out on the platform. An accompanying survey found that heavy users (using TikTok for over three hours a day) had a significantly more positive view of CCP’s human rights record than non-users.”
The investigation discovered that heavy TikTok users had a 49% increase in positive assessments of CCP’s human rights record compared to non-TikTok users. In contrast, the use of YouTube and Instagram did not show a significant relationship with users’ views on CCP’s human rights record.
Unlike YouTube and Instagram, TikTok usage significantly influenced users’ “positive perceptions” of using “Tiananmen Square” as a tourist destination. Heavy TikTok users demonstrated a 41% higher agreement on “Tiananmen Square as a famous tourist spot” compared to non-users.
Finkelstein wrote: “This manipulation extends not only to the availability of content but also to psychological manipulation, particularly affecting Generation Z users.”
The report stated, “These insidious strategies have far-reaching implications, effectively shaping the viewpoints of their target audience – young users. These users seem to unwittingly absorb these biased narratives, leading to distorted understandings of global important issues.”
The report confirmed that “TikTok’s algorithm actively suppresses content criticizing the CCP while strengthening pro-CCP propaganda and promoting irrelevant content that diverts attention.”
Although the current findings cannot prove manipulation by the Chinese government, they provide “convincing compelling evidence” of TikTok’s secretive content manipulation.
Last December, after the initial report by the NCRI, TikTok shut down its tag measurement feature, hindering researchers from replicating the study.
To continue this research, investigators collected “user journey” data through organic accounts to simulate the social media experience of American teenagers.
The report mentioned that the first report only revealed the possibility of TikTok content manipulation but did not delve into specific auditing algorithms or operations, and added that this new report addresses these shortcomings.
A TikTok spokesperson refuted the findings of this report, claiming that creating new accounts and searching for these keywords does not reflect the actual user experience.
In April this year, U.S. President Joe Biden signed a law requiring ByteDance to sell its TikTok U.S. assets by January 19, 2025, or face a nationwide ban.
Subsequently, TikTok sued the U.S. government to overturn the law, arguing that Congress had no evidence to substantiate its claim that the app threatens national security.
The Network Contagion Research Institute (NCRI) at Rutgers, The State University of New Jersey, is an independent nonprofit organization composed of political scientists, security experts, and research analysts. The organization is funded by Rutgers University, the UK government, and private charitable institutions.