A forceful warning against wasteful government investments intended for local authorities appeared on Sunday in a central-level newspaper dedicated to economic policy. Though the source of the message was masked in one sense by an official pen name, its import was nonetheless unmistakable — revealing a key fact about how China communicates policy priorities.
The commentary, published in the Economic Daily (经济日报), a newspaper directly under the State Council, highlighted concerning evidence of declining investment efficiency as Beijing pushes to stimulate domestic demand. The article noted that China’s incremental capital output ratio — which measures how much new investment is required to produce each additional unit of GDP growth — has deteriorated from 2.84 in 2008 to 9.44 in 2023, higher numbers indicating worsening efficiency.
These numbers suggest that China now requires more than three times the capital investment to generate the same economic output as 15 years ago. This, the commentary suggests, pointing a finger at local governments, indicates serious structural inefficiencies in how capital is being deployed across China’s economy.
The article appeared under the byline “Jin Guanping” (金观平), a homophone for “Economic Daily’s observational commentary” (经济日报观察评论), signaling that this represents an institutional position rather than an individual author’s view. In all likelihood, “Jin Guanping” commentaries are penned by a “writing group,” or xiezuozu (写作组) at the newspaper, with senior editors and possibly officials reviewing and approving the final text before publication.
“Some localities blindly launched vanity projects to pursue short-term political achievements, causing large amounts of capital to flow into inefficient projects,” the commentary said. It castigated local governments for pursuing what it called “grandiose plans” (大手笔) that had resulted in “massive debts” (大笔债) — and, in the end, yielded little in terms of economic benefits.
Finance Minister Lan Fo’an (蓝佛安) revealed in November that local governments had accumulated 14.3 trillion RMB (1.97 trillion dollars) in hidden debts by the end of 2023, according to Hong Kong’s South China Morning Post.
What’s in a Pen Name?
The “Jin Guanping” byline has frequently been used for commentaries at the Economic Daily, pointing to consensus viewpoints in the central government. On average, the byline appeared for 13.3 articles per month in the newspaper over the past year. What distinguishes Sunday’s commentary, however, is its cautionary tone. It directly criticizes wasteful projects with terms like “grand gestures” (大手笔) and “large-scale debt” (大笔债), and it warns that such cases bear “profound lessons” (教训深刻).
This practice of using “homophonous pen names” is an internal coding system within China’s official state media. The pen names, even as they seem to disguise the source of the messaging, often work as signals indicating the institutional weight behind particular messages. As we have previously noted at the China Media Project, these propaganda labels “form an internal system of not-so-secret codes by which those in positions of power, both departments and individuals, can voice their official positions and put their stamp on a course or policy.”
The practice of using official pen names dates back decades in China. During the Cultural Revolution, radical factions used pen names to signal political attacks. In more recent times, pen names have regularly appeared in Party-state media such as the flagship People’s Daily. For example, the pen name “Zhong Sheng” (钟声) — a homophone for “Voice of China” — is used to mark important commentaries on international affairs where the leadership wishes to register its view, often critical, of foreign countries without direct attribution. The pen name “Guo Jiping” (国纪平), a homophone of “important international commentary,” is used on occasion to mark international affairs commentaries that represent the central CCP consensus.
By parsing these coded bylines, observers can better understand not just what is being communicated, but who might be communicating it and how seriously the message should be regarded within China’s complex political-economic system.
The appearance of this stern warning under the “Jin Guanping” byline at this particular moment suggests heightened central government concern about local investment practices as China struggles to revive economic growth.
According to rare reports today from Chinese media, an environmental crisis is unfolding along a stretch of the Leishuei River in Hunan province that impacts the prefectural city of Chenzhou (郴州), home to more than four million people. Abnormal concentrations of thallium — a highly toxic, colorless heavy metal that causes organ damage and cancer through water contamination — have reportedly prompted the city to activate a Level IV emergency response, and residents are stockpiling drinking water.
In neighboring Guangdong province, the Southern Metropolis Daily (南方都市报), a commercial newspaper published under the state-run Nanfang Daily Group, splashed the crisis across its front page today, with the headline: “Thallium Abnormality in Hunan’s Leishuei River.”
The front page of today’s edition of Guangzhou’s Southern Metropolis Daily.
According to reports from both the Southern Metropolis Daily and Caixin Media, the crisis began nearly a week ago, on March 16, as automatic monitoring stations along a section of the river between the cities of Chenzhou and Hengyang, population 6.6 million, showed abnormal thallium levels, “causing trans-municipal pollution and threatening downstream water safety” (造成跨市污染，威胁下游饮水安全).
Jimu News, an online official outlet from Hubei province, immediately to the north of Hunan, reports that both Chenzhou and Leiyang cities have established emergency command centers to address abnormalities in local water quality in the Leishuei basin. Local officials have insisted that drinking water remains safe in the area impacted by the abnormal readings. However, Shanghai’s The Papersaid in a report this afternoon, adding wider context to the breaking story, that abnormal thallium concentrations had been detected in 17 out of 22 drinking water sources along the Xiangjiang River in Hunan province, a separate basin, since 2020 — pointing potentially to wider and more longstanding public health risks.
It was only yesterday that that local government in Chenzhou finally acknowledged publicly through its government website what local officials had been responding to for a week: “Water quality in some sections of the Leishuei basin has shown abnormalities.” This public statement came seven days after automatic monitoring stations first detected abnormal thallium levels on March 16, and six days after Yongxing County activated its Level IV environmental emergency response, the Southern Metropolis Daily reported.
Caixin Media reports on the Hunan thallium story through its English account on X this afternoon.
Hundreds of gigabytes of data lurking on an unsecured server in China linked to Baidu, one of the country’s largest search engines and a major player in the fast-developing field of artificial intelligence (AI), offer a rare glimpse into how the government is likely directing tech giants to categorize data with the use of AI large language models (LLMs) — all to supercharge the monitoring and control of content in cyberspace.
First uncovered by Marc Hofer of the NetAskari newsletter, the data is essentially a reservoir of articles that require labeling, each article in the dataset containing a repeated instruction to prompt the LLM in its work: “As a meticulous and serious data annotator for public sentiment management, you must fully analyse article content and determine the category in which it belongs,” the prompt reads. “The ultimate goal is to filter the information for use in public opinion monitoring services.”
In this case, “public opinion monitoring,” or yuqing jiance (舆情监测), refers broadly to the systematic surveillance of online discourse in order to track, analyze, and ultimately control public sentiment about sensitive topics. For social media platforms and content providers in China, complying with the public opinion monitoring demands of the Chinese government is a herculean effort for which many firms employ thousands of people — or even tens of thousands — at their own cost. This leaked dataset, of which CMP has analyzed just a small portion, suggests that this once-human labor is increasingly being automated through AI to streamline “public opinion monitoring and management services,” known generally as yuqing yewu (舆情业务).
Extract of the dataset, an instruction to classify a piece of data according to 38 described categories
What does the dataset tell us?
First, it reveals a sophisticated classification system with 38 distinct categories, running from more mundane topics like “culture” and “sports” to more politically sensitive ones. Tellingly, the three categories marked as “highest priority” in the dataset align distinctly with state interests as opposed to commercial ones. Topping the list is “information related to the military field,” followed by “social developments” (社会动态) and “current affairs developments” (时政动态). This prioritization underscores how private tech companies like Baidu — though it could not be confirmed as the source of this dataset — are being enlisted in the Party-state’s comprehensive effort to monitor and shape online discourse.
The scope of this monitoring operation is reflected in the sheer volume of data — hundreds of gigabytes found on an unsecured server. While many questions about the dataset remain unanswered, it provides unprecedented insight into how Chinese authorities are leveraging cutting-edge AI technology to extend and refine their control over the information environment, pressing the country’s powerful tech companies to serve as instruments of state surveillance.
Weathermen and Forecasters
To understand the significance of the “public opinion monitoring” this dataset supports, we must turn the clock back to 2007, the year that saw the rise of microblogging platforms in China, fueling real-time engagement with current affairs by millions of internet users across the country. In comparison to today, China’s internet at that time was still relatively untamed. That year, one of a number of major controversies erupting in cyberspace was what eventually became known as the “Shanxi Brick Kiln Incident” (黑砖窑事件) — a “mass catharsis of public anger,” as Guangzhou’s Southern Metropolis Daily newspaper dubbed it.
The scandal, exposed only through the dogged determination of concerned parents who scoured the countryside for their missing children, revealed that over 400 migrant workers, including children, had been held in slave-like conditions at a brick kiln complex in Shanxi province — a situation one court judge candidly admitted in the scandal’s aftermath was “an ulcer on socialist China.” As news and outrage spread virally online in June 2007, it ballooned beyond the capacity of the state’s information controls. Party-state officials witnessed firsthand the power of the internet to mobilize public sentiment — and, potentially, threaten social and political stability.
Screenshot of a report on China Central Television showing enslaved workers liberated from kilns in Shanxi. SOURCE: CCTV.
This watershed case fundamentally transformed the leadership’s approach to managing online discourse. What began as a horrific human rights abuse exposed through citizen journalism became the catalyst for what would evolve into a sophisticated public opinion monitoring apparatus with national reach, and a booming industry in public opinion measurement and response.
By 2008, the “Shanxi Brick Kiln Incident” had kickstarted the “online public opinion monitoring service industry” (网络舆情服务行业), an entire ecosystem of information centers set up by state media (like the People’s Daily and Xinhua News Agency), as well as private tech enterprises and universities. Analysts employed in this growing industry were tasked with collecting online information and spotting trending narratives that might pose a threat to whomever was paying for the research — in many cases provincial and local government clients, but also corporate brands.
While the primary motivation was to forestall social and political turmoil, serving the public opinion control objectives of the leadership, the commercial applications of control were quickly apparent. Five year laters, Guangzhou’s Southern Weekly (南方周末) newspaper would report on the “big business” of helping China’s leaders “read the internet,” with revenues from related business at People’s Daily Online, a branch of the CCP’s own People’s Daily, set to break 100 million yuan, or 16 million dollars. According to the paper, 57 percent of public opinion monitoring clients at the time were local governments.
“For government departments at all levels, the need to understand online public opinion has become increasingly urgent,” the Southern Weekly captioned this image in 2013. The chart shows public opinion incidents peaking in June, November and December each year. Local governments account for 57 percent of clients at the time. SOURCE: Southern Weekly.
“If online public opinion is an important ‘thermometer’ and ‘barometer’ for understanding social conditions and public opinion,” the founder and director of the People’s Daily Online Public Opinion Monitoring Center (人民网舆情监测室), Zhu Huaxin (祝华新), said at the time, “then public opinion analysts are ‘weathermen’ and ‘forecasters.’”
The job of China’s public opinion forecasters and weathermen has evolved over the past 18 years. In 2016, as the industry neared the end of its first decade, and as online public opinion continued to move faster than analysts could manage, China Social Sciences Today (中国社会科学报), a journal under the government’s State Council, urged the system to upgrade by applying “big data” (大数据). Over the past decade, automating public opinion services and cutting down on costs has been the goal in the evolving business of managing public opinion. Today, the entire system is now being supercharged by AI.
Those gigabytes of data lurking on an unsecured Baidu server offer us a closer look at how the public opinion monitoring work of AI is being organized.
A Cog in the Machine
What exactly does the prompt in this dataset do? When copy-pasted along with a news article into Chinese large language models like Baidu’s Ernie Bot (文心一言) or DeepSeek, the prompt instructs the AI to classify the article into one of the 38 predefined categories. The LLM then outputs this classification in json format—a structured data format that makes the information easily readable by other computer systems.
This classification process is part of what’s known as “data labeling” (数据标注), a crucial step in training AI models where information is tagged with descriptive metadata. The more precisely data is labeled, the more effectively AI systems can analyze it. Data labeling has become so important in China that the National Development and Reform Commission released guidelines late last year specifically addressing this emerging industry.
When the prompt is put to Baidu’s Ernie Bot, it provides one of the listed classifiers as an output, in code format.
The dataset strongly suggests that Baidu is using AI to automate what was once done manually by tens of thousands of human content reviewers, with varying levels of automation. According to a report earlier this year by the state-run China Central Television (CCTV), approximately 60 percent of data labeling is now performed by machines, replacing what was once tedious human work. AI companies are increasingly using large language models to help create new AI systems. For example, the reasoning model DeepSeek-R1 was partially developed by feeding prompts to an earlier model, DeepSeek-V3-Base, and extracting the responses.
Monitoring and Manipulation
What can we learn from the three “public opinion related” categories that Baidu’s dataset identifies as “most important”? While we couldn’t find official regulations from the Cyberspace Administration of China (CAC) specifically using these three categories, the content in these classifications reveals what the Chinese government considers most critical to monitor.
A report in 2010 reviews what at the time was the short history of the public opinion monitoring profession.
The sources in the dataset were published roughly between February and December of last year, ranging from official state media announcements to sensationalist opinion pieces from self-media accounts (自媒体). Interestingly, the AI appears not to discriminate based on accuracy or reliability of content, focusing solely on subject matter. Some content could not be clearly categorized. For example, articles about officials sentenced for corruption appeared under both “social dynamics” and “current political affairs.”
Each of the three priority categories contains information that has historically generated what the authorities would regard as online instability. “Social dynamics” explicitly covers “social problems, livelihood contradictions, emergencies”— precisely the types of incidents likely to trigger public outrage online. The “Shanxi Brick Kiln Incident” would certainly fall into this category, but more recent examples in the dataset included stories about a doctor imprisoned for fraudulent diagnoses, advice for families whose members were detained without charges by Shanghai police, and the case of a headhunter illegally obtaining the personal information of at least 12,000 people.
Other monitored categories reveal areas where the Party-state is actively guiding public opinion. “Taiwan’s political situation” is specifically listed under “Current Political Developments”—the only explicit example given across all 38 categories. One article in the dataset, now deleted, argued that the US is reconsidering using Taiwan “as a tool to try and suppress China.” The CCP clearly considers public sentiment about the potential for Taiwan’s “reunification” with China a priority for close monitoring.
Similarly, military information is closely watched. Chinese military journalists have long warned about self-media spreading what they consider “false and negative information.” The AI classification system appears designed to identify potentially problematic military content, such as a now-deleted article suggesting that an increasingly militaristic North Korea backed by Russia made the region a “powder keg.” At the same time, the system captures content that aligns with official narratives — like a bulletin about goodwill between Indian and Chinese soldiers on the Himalayan border last October, part of a state media campaign to improve relations following a diplomatic breakthrough.
The exact purpose of this dataset remains unclear. Were these classifications developed internally by Baidu — or were they mandated by state regulators? Nevertheless, the unsecured data offers a glimpse into the inner workings of China’s AI content dragnet. What was once a labor-intensive system requiring thousands of human censors is rapidly evolving, thanks to the possibilities of AI, into an automated surveillance machine capable of processing and categorizing massive volumes of online content.
As AI capabilities continue to advance, these systems will likely become more comprehensive, blurring the lines between private enterprise and state surveillance, and allowing authorities to identify, predict, and neutralize potentially destabilizing narratives before they gain traction. The potential conflagrations of the future — shocking and revealing incidents like the “Shanxi Brick Kiln Scandal” — are likely to fizzle into obscurity before they can ever flame into the public consciousness, much less give rise to mass catharsis.
In a city where Buddhist rock carvings have gazed over the Yellow River valley for more than a millennium, local officials are now on their own sacred mission — to join the Chinese state in amplifying its message to the rest of the world.
On Wednesday this week, Chinese Communist Party officials gathered at the offices of the official Luoyang Daily (洛阳日报) to inaugurate the city’s new Luoyang International Communication Center (洛阳国际传播中心), the latest addition to China’s expanding network of provincial media hubs designed to “tell China’s story” to international audiences. For Luoyang — an industrial center in central China’s Henan province that served as imperial capital during multiple dynasties — the new center aligns ancient heritage with contemporary propaganda objectives.
The state-run China Daily newspaper reported that the center will “integrate government, media, academic institutions and social resources to build a comprehensive, multi-level international communication system.” The paper, which is itself a key external propaganda organ published by the Information Office of the State Council — essentially the same office as the CCP’s Central Propaganda Department — called the initiative “an important platform for Luoyang to promote high-quality development of international communication and enhance international discourse power” (提升国际话语权).
Since 2018, and accelerating since 2023, the Chinese leadership under Xi Jinping has called on provincial and city-level governments to join the national push to raise the country’s “discourse power” — hoping to close what it sees as a gap with the influence of the West — by launching local “international communication centers,” or ICCs. These are meant to leverage the resources of local media groups and propaganda offices, along with partnerships with universities and other entities, to promote Chinese cultural, economic and political influence.
A new crop of official ambassadors for Luoyang. Wild guess: These are foreign students recruited for a photo opp. SOURCE: China Daily.
Alongside the launch of the Luoyang center, officials announced the formation of the “Luoyang International Communication Alliance” (洛阳国际传播联盟), which according to official reports comprises more than 100 members from various sectors with overseas experience. The Luoyang ICC has also signed cooperation agreements with four local universities, including Henan University of Science and Technology — presumably to develop content and recruit talent for international messaging efforts. Foreigners from a number of countries, including the United States, the United Kingdom, Russia, Belarus, South Korea, and Morocco, were present at the launch ceremony, promoted as “Luoyang International Promotion Ambassadors” (洛阳国际推荐官).
Whether these provincial centers will significantly influence international opinion remains to be seen. State media reports, including from China Daily, claim that since preliminary operations of the Luoyang ICC began in December 2024, the center has “actively expanded overseas communication channels” and produced more than 600 articles and videos — content that reportedly attracted attention from China’s own Foreign Ministry. Aside from occasional announcements about the ICC’s formation posted on Instagram and Facebook by the state-run China News Service, however, we could find no evidence of any meaningful international impact.
Shanghai’s Fudan University (复旦大学) is one of China’s most prestigious universities, with a raison d’etre unchanged, it claims, since the institution was founded in 1905: improving China’s position in the world through education. As artificial intelligence takes the world by storm — and becomes a crucial priority from top to bottom in China — the means of achieving that mission is changing, according to the university’s president, Jin Li (金力).
On February 25, Jin announced that Fudan would drastically reduce its course offerings in the humanities, instead focusing on AI training. In an interview with Guangzhou’s Southern Weekly (南方周末) on March 6, Jin said the university wanted to cultivate students that “can cope with the uncertainty of the future.” For Li, cutting the liberal arts cohort by as much as 20 percent is a social necessity. As he asked rhetorically in the interview: “How many liberal arts undergraduates will be needed in the current era?” (当前时代需要多少文科本科生?).
At present, courses related to artificial intelligence at Fudan are at 116 — and counting. And the university isn’t alone in downsizing the arts. Combing through Ministry of Education statistics on university courses cancelled in 2024, the commercial newspaper Southern Metropolis Daily (南方都市报) noted that the majority were for liberal arts degrees, with some universities even abolishing their humanities colleges altogether.
Limiting the humanities comes at a time of broader upheaval in higher education within China. In 2023, the Ministry of Education issued a reform plan ordering that by this year, 20 percent of university courses must be adjusted,with new course offerings introduced to “adapt to new technologies.” According to the plan, majors “not suitable for social and economic development” should be eliminated altogether.
Limiting the humanities comes at a time of broader upheaval in higher education within China.
AI is almost certainly foremost in the ministry’s mind as it considers plans for the overhaul of education. The country’s “AI+” campaign, introduced during last year’s National People’s Congress, pegs the new technology as key to China’s future development — the source of “new productive forces” (新质生产力) that will rejuvenate the economy. As such, some universities are expanding their offerings in AI courses, making AI literacy classes compulsory for students, and allowing a lax approach to using AI in research. Tianjin University, for example, has decreed students can use AI-generated content for up to 40 percent of a graduation thesis. But that raises the obvious question: if a machine writes 40 percent of your paper, have you really only learned 60 percent of the content?
Since 2023, there have been increasingly lively debates — and much hand-wringing — about the ethics and limitations of AI use in higher education. In China, it seems, it is full steam ahead.
In a stark illustration of media corruption in China, a prominent financial blogger named Kou (寇某), who ran a pair of prominent consumer-related social media channels, was sentenced on March 12 to more than 10 years in prison for “news extortion” (新闻敲诈). The case, which unfolded between 2022 and 2023, highlights a chronic problem in the country’s controlled media environment. By fabricating false news stories about companies and threatening reputational damage, he systematically extorted over 700,000 yuan through manipulative tactics that exploit the vulnerabilities of China’s information ecosystem, according to a report by the Xinmin Evening News (新民晚报), a paper under Shanghai’s state-run SMG.
The recent reporting stems from the completion of Kou’s legal case and ongoing efforts, officials say, to clean up the online information environment — serving as a warning to other potential offenders. The channels in question, “Consumer Financial Channel” (消费金融频道) and “Payment Encyclopedia” (支付百科), were previously included in a 2023 list of cautionary cases issued by the Cyberspace Administration of China (CAC), the country’s top internet control body. That list accused the channels of exploiting the principle of “public opinion supervision” (輿論監督) — a term associated with media monitoring and even investigative reporting — to commercially exploit companies.
Kou’s “Consumer Financial Channel” was also profiled last year in a feature story from Shanghai’s The Paper (澎湃). In addition to Kou’s case, that story detailed a 2019 case in which a young couple in Shanghai illegally established 41 fake news websites, systematically publishing over 100,000 negative articles about companies and individuals across 20 provinces. By charging 500 yuan per negative article and targeting thousands of businesses, the couple earned over 1 million yuan in less than a year through their online extortion scheme.
What are the vulnerabilities of China’s information ecosystem to which the Xinmin Evening News report referred?
One crucial underlying issue is the state’s control of the press environment itself, which has led to rent-seeking behavior by powerful state media journalists who can threaten exposure of alleged malfeasance and demand “hush money” (封口費). Importantly, cases of news extortion over the past two decades in China have often implicated powerful state-run media as well as outright fraudsters. The most notorious example is the 2002 “Gold Nugget Case,” where eleven reporters, including four from Xinhua News Agency, accepted cash and gold nuggets to cover up a mining explosion that killed 37 workers, highlighting the deep-rooted corruption within China’s media landscape.
Generation Z has now become the primary force among China’s growing ranks of China’s online content moderators, who number in the tens of thousands. Their physical stamina means they generally fare better with the intense demands of the job and can stay up late to respond quickly to issues with sensitive content. They now handle the bulk of the work for major internet platforms when it comes to content moderation.
Also known as the “Internet Generation,” Generation Z also includes those born after 1995, who from birth have known only a world with the internet, which arrived in China in 1994. They are not just familiar with the internet, mobile technology, and smart applications, but have also grown up alongside the Great Firewall, which since 1996 has cut them off from the global internet.
Data shows that ByteDance, now firmly established as China’s leading internet company, employs more than 100,000 people, of which more than 20 percent work as content moderators. At Bilibili, a popular video-sharing platform that targets younger audiences, this proportion exceeds 27 percent.
Headquartered in Beijing, ByteDance Ltd. employs more than 100,000 people, of which nearly one-fifth are content moderators. SOURCE: Bytedance.
China, with more than one billion internet users, represents 19 percent of global users of the internet. However, even as the number of content moderators has been on the rise, the number of Chinese web pages has fallen by 70 percent in the last ten years, and the number of Chinese websites has declined by 30 percent in the last five years. While hundreds of millions of Chinese internet users divide their time among dominant platforms like Baidu, Weibo, WeChat, and Douyin, strict content censorship has effectively confined them within an information firewall of “China-exclusive” (中国特供) internet content.
In creating this situation, online content moderators, who serve as the ultimate enforcers of state policy, have played an indispensable role.
Born in 1997, Chen Lijia in 2020 became a content moderator at a prominent mainland Chinese internet company that operates the country’s largest search engine. The company’s founder has publicly stated that its user base exceeds one billion. And all search queries and content published by these one billion users must pass through the massive moderation team that includes Chen. Among the various types of content moderation, politically sensitive material receives the highest priority.
Chen Lijia is candid about his reasons for becoming a content moderator. He simply wants to earn a living. But Chen and his colleagues have spent their entire lives within the confines of China’s Great Firewall. So how can they effectively judge which information should be censored?
“The company holds regular training sessions to tell us what kind of information we need to remove,” Chen says. “But they never explain why. We’re just treated as tools.”
It is precisely because he understands his position as a “tool” that Chen sees his work as unrelated to justice or morality — but merely about survival. Facing a deteriorating economy and an increasingly competitive society, this is how many Gen-Z content moderators genuinely feel as they routinely delete posts and suspend accounts containing politically incorrect content.
What Do Content Moderators Do?
Q: What kind of work do content moderators at your company mainly do?
A: Mainly image and text moderation (图文审核), the content dealing mostly with three areas —political security, violence, and pornography. Beyond these, we also need to moderate content here and there related to gambling, drugs, and illegal advertisements.
Q: What kind of content is considered illegal advertising?
A: We mainly remove advertisements from companies that don’t have partnerships with us. Advertising is one of the main sources of revenue for our company, and if we don’t control these unauthorized ads, the company can’t make money. That’s why content moderators receive a whitelist from the company, and any advertisements not on that list need to be removed.
Q: Among the content that needs moderation, which category has the highest priority?
A: Political security, definitely.
Q: What are the standards for reviewing political content? For example, what type of content passes review and what doesn’t?
A: The standards are actually similar to those used in domestic news reporting. Any content that can’t be reported in the news media is content that can’t pass our review. The most obvious examples are content related to high-ranking political officials and their families, or content about the Communist Party’s negative historical events. We absolutely must remove all of that stuff. Every year around June 4, people share information about the June Fourth Incident, and we have to remove all of it. As June 4 approaches, we even have to work overtime to manage everything.
Q: So during special time periods like June 4 and National Day on October 1, you moderators need to be extra vigilant, right?
A: Absolutely! Because during these time periods, there are always people who for whatever reason want to refresh public memory. As these dates approach, they like to post different things. From our perspective, June 4 has essentially become an unofficial folk holiday (民间节日). During these sensitive periods, the review rules and strategies temporarily change. Normally we use a “post first, review later” approach, but during sensitive periods, it switches to “review first, then post.”
Q: Besides changing from “post first, review later” to “review first, then post,” are there any other adjustments to strategies or rules?
A: No one has ever explicitly explained these things to us. Content moderators are just the implementation level. In my years at the company, they (the full-time employees) have never allowed me access to their moderation policies. We simply have to do whatever they tell us to do.
Q: Are there changes at the implementation level during those times?
A: Things just become stricter and more intense. Everyone has to work overtime then, and we also have to work in rotating shifts, including overnight shifts.
Q: Does the company explicitly tell you what content must be removed?
A: They don’t state it explicitly. They just give us vague examples. They tell us to be vigilant and ensure certain content doesn’t get through. But they never clearly explain to us exactly what these things are about.
Q: What sort of things did they test for when you interviewed for the political content moderator position?
A: They would introduce certain sensitive political information, and then ask me how I’d handle it. I’m pretty familiar with such things, and I even knew more about them than the examiner. On the June Fourth Incident, for example, after a little training, everyone knew what it was. But when there were terms like “Tiananmen bullet holes” (天安门子弹孔) — aside from me, no one else was aware of them. This is what sets apart those capable of doing this kind of work. It’s crucial to be familiar with the vocabularies that derive from such incidents — because many people use things like symbols and homophones to disguise what they’re saying. For example, the term “public square defender” (广场卫士) is an expression that comes from June Fourth. I’m able to discern it with just a glance. Any good content moderator must possess this degree of political sensitivity.
Q: Is the “People’s Liberation Army” something that can be mentioned?
A: Around sensitive dates, it’s totally forbidden. Even comments praising the government and the PLA for restoring order are off limits. The point of this censorship is not to drive public opinion in any certain direction. It’s beside the point whether something is positive or negative. The point is for censors to obliterate the event so that the public completely forgets about it.
Q: What other types of content do they bring up during training?
A: The Tiananmen bullet hole thing — they didn’t need to train me on that. I’ve known about that for quite some time. Even my bosses weren’t aware of it until I spread the word around. I’m really diligent about my job.
Q: So those who can get over the Great Firewall and have a better understanding of politics can perform political content moderation better than others.
A: Of course. So I could advance as a quality inspector faster than others.
Q: How many training sessions have you taken part in with your current employer?
A: Maybe seven or so. The one that made the deepest impression was on labor camps in Xinjiang. The sessions were totally shallow, and they really taught you nothing. The company’s point was just to make you aware — that was it. As for any of the specifics, we didn’t need to remember those. I think the company really preferred that we forget everything after the training.
Q: They don’t tell you why certain things are politically sensitive? Do you ever look into it yourself out of curiosity?
A: The company doesn’t need us to know anything in too much detail. In fact, AI will already have blocked out a lot of content through keyword filtering — and that content might have been much more detailed. But even we content moderators can’t see that stuff. I’m not someone who would look into it, because I’ve understood it for a long time. For things I don’t understand, I’ll scale the wall and check it out on the foreign internet. When you understand more, sometimes during the manual review process, I discover certain keywords, and then I report them to leadership. Afterward, these keywords get added to the computer program, and then the AI conducts automatic review and blocking.
The one that made the deepest impression was on labor camps in Xinjiang . . . . I think the company really preferred that we forget everything after the training.
Q: What keywords have you reported up the chain of command?
A: For example, “egg hole” (蛋孔) — that’s “egg” as in a chicken egg — and “dan fried rice” (旦炒饭), that’s the “dan” as in “Satan,”, and so on. Also combinations of the characters for “new” and “frontier” [as in the name of “Xinjiang”] with various homophones. There are many of them, but they’re all quite trivial. [NOTE: These are coded references mocking the death of Mao Zedong’s eldest son, Mao Anying, during the Korean War. According to internet lore, Mao’s son was killed by American bombing after revealing his position by cooking egg fried rice on the battlefield.]
Q: How did they conduct the training about the labor concentration camps in Xinjiang?
A: They just had us watch a documentary. After we finished watching, that was it — nothing else was said. The company’s goal wasn’t to teach us the truth but to make us delete content. Different intentions lead to different approaches. They just wanted reviewers like us, mere cogs in the machine, to recognize this issue as sensitive and simply delete relevant keywords whenever we encountered them. They didn’t care about our thoughts or feelings on the matter.
Q: So you guys just need to know that “Xinjiang” and “concentration camps” are sensitive terms, and then as long as you delete them immediately upon seeing them that’s enough.
A: No. The term “concentration camp” isn’t sensitive on its own. Only when it’s combined with “Xinjiang” does it become sensitive. That’s exactly why human review is necessary. Machines can only moderate obvious content, while humans can detect subtleties.
For example, my bosses and I once had a dispute about content asking, “Why didn’t Chiang Kai-shek assassinate Mao Zedong when he was in Chongqing?” I thought this sentence was normal, but our leaders demanded its deletion. I resisted but eventually deleted it without understanding why. Later I realized that Mao Zedong is considered a great person — so how could anyone suggest assassinating him? From the state’s perspective, that statement is clearly politically incorrect. So to be effective human reviewers rather than machines, we need to better understand such implied meanings.
Q: What kind of content have you all been trained on?
A: We’ve had training about evil cults, like Eastern Lightning. But it was just a PowerPoint presentation. No photos were permitted. As soon as it finished, the company quickly took everything away. All of the company’s training is secretive like that. My impression is that they need us to know about these things to do our job properly, but at the same time, they’re afraid of us knowing too much. They’d prefer we forget everything after executing their instructions. It’s really messed up.
Q: In terms of evil cults, what content had you already been aware of beforehand?
A: I really didn’t know about this before. And it was precisely because I wasn’t aware and didn’t understand that I didn’t feel any guilt about deleting posts. Even when deleting content about June Fourth, I felt no guilt. These things just seemed so remote to me.
Guilt and Morality
Q: Is there anything that gives rise to a feeling of guilt for you?
A: Yes. For example, content about the Covid epidemic, and about the floods in Zhengzhou. There was also this piece called “Ten Days in Chang’an” (长安十日), written by someone called Jiang Xue (江雪). Deleting those things made me feel guilty. But these were very obvious things [that demanded deletion]. If I didn’t delete them, someone else would have. They were too blatant. I couldn’t have let them pass even if I wanted to. [NOTE: Jiang Xue is a well-known non-fiction writer and former journalist from Xi’an.]
Food packets being delivered to residents in the city of Xi’an under quarantine in 2020. SOURCE: China Digital Times.
Q: How does your sense of guilt manifest?
A: It makes me want to quit. But people around me console me by saying that if I hadn’t censored it, someone else would have. My friends also ask what I would do instead. Everyone knows survival comes first. Morality only exists after survival is taken care of.
Even so, 2021 was an extremely difficult year for me. The Xi’an incident and the Zhengzhou floods happened that year. That’s when my guilt peaked. I really wanted to leave the company and never work as a political censor again, but I couldn’t find any good opportunities. So political censorship has become something quite contradictory for me. On the one hand, I’m skilled at it and it puts food on the table. On the other hand, it’s . . . really painful. My guilt becomes especially intense when I’ve personally experienced the events I’m censoring.
Q: Does guilt leave you politically depressed (政治性抑郁)?
A: I feel more like my conscience is condemning me. I really don’t like the term “politically depressed.” I’ve rarely heard it used before. I prefer to use the word “conscience” to express how I feel. This word better reveals my state of mind than “politically depressed.” [NOTE: “Political depression” refers to a form of depression triggered by political events or circumstances. It differs from typical clinical depression by manifesting when individuals feel powerless over societal or governmental situations.]
Q: After the condemnation of your conscience, what has your response been in real life?
“I feel more like my conscience is condemning me.”
A: I’ve started to really love chatting with my friends.
Q: Are you hoping for some external support from your friends’ words that might allow you to keep doing this job? Or is it that they’ll give a kind of “legitimacy” or approval to your work?
A: I can’t say there’s none of that, but I don’t think my intention is so clear.
Q: So what do you think your intention is in seeking out friends to talk to?
A: I just want to vent and find like-minded people. Of course I also want to ask them if there’s another path I can take. After all, this road [of being a content moderator] isn’t sustainable in the long run.
Q: In your eyes, what would make sense in the long run for you?
A: I have always wanted to use this job as a springboard that could help me jump to a civil service position in a staff role to an official. I chose to join a public opinion monitoring company because I wanted to build my skills in resolving public opinion risks — essentially crisis and threat management abilities — and then move on from there to a better platform. But such opportunities are rare and hard to come by. What I’ve always really wanted to do is to work at an organization like the State Council Development Research Center, where I could be an aide to a higher-ranking official, advising them on how to resolve crises. That’s the job I most want to do.
Q: But for this kind of job you need an excellent academic background and comprehensive skill sets.
A: The fact that I don’t have an excellent academic background is exactly why I wanted to enter a public opinion monitoring company, do political content moderation, and even excel at it. I thought that once I’ve developed certain capabilities, it might be easier for people to take notice of me. That’s been my personal aspiration.
Q: So now that you’ve achieved a level of technical skill in political content moderation, do you think your personal capabilities have improved?
A: There’s been no qualitative improvement, only some quantitative, or self-perceived improvement.
Q: What improvements have you noticed in yourself?
A: I think because I’ve seen so many bits of information, I’m now able to use alternative methods to resolve issues. But my perspective has always been that of an ordinary person. I’ve never considered problems from the viewpoint of a high-level official. That’s the biggest challenge for an aide. I think it’s also my biggest obstacle to becoming an aide, because the perspective is a completely different one.
Q: But have you never thought that, in doing content moderation, you also aren’t at the vantage point of a normal person or even of yourself? Aren’t you always having the vantage point of someone in control?
A: That’s right, but this vantage is just one aspect of what it means to put out fires — meaning that by putting out fires you can calm things down. But my position isn’t completely that of someone in control because beyond implementation, I really have no idea how their strategy is designed, or how those making the strategy communicate with the government news office.
These factors have severely limited my growth. I can’t even attend those meetings. I can only guess how they communicate with leadership. I used to believe that after doing a good job with content moderation, I would get opportunities to advance and increase my value personally. Eventually, though, I found I couldn’t even become a formal employee.
Q: Do you feel your situation is tragic? And if you could step back and view it objectively, how would you evaluate the profession of content moderator?
A: It really is tragic. Looking at content moderators as a group, I can only say these are people working to eat, to survive. It’s simply about putting food on the table, nothing more than that. As for questions about “righteousness” — nobody really examines that deeply. When you can’t even feed yourself, what righteousness is there to speak of? You could call it the helplessness or sorrow of people at the bottom, because for people at the bottom, there aren’t many choices.
So I’ve never accepted moral criticism of this job. Shouldn’t moral criticism be directed at the middle class? Why throw it at those at the bottom? What I want to ask now is: who designed the position of content moderator? We’re just doing labor. This definitely isn’t a problem with the job itself, but with the people who created it. So the criticism should be directed at the source of the problem, not at those of us who are just following orders. After all, there’s an endless supply of cogs. If not us, there would be others.
I think there’s a good saying — that the rights and position you have determine what responsibilities you should bear. You can’t transfer too much social responsibility to ordinary people. Ordinary people already struggle to survive. They can’t shoulder such heavy expectations.
Q: So do you believe that there is some way to address the source of this problem?
A: No. And there are so many people in society who need to eat. If you really don’t let people do this job anymore, how are they supposed to survive? People might say, “Oh, it’s just a job.” But if this job doesn’t exist, do you have another for me? Do you really? So not resolving the issue also has its benefits. Ultimately, you can only address moral problems after solving the problem of having enough to eat. If you can’t resolve even basic subsistence issues, we’re nothing more than beasts.
So I really dislike certain experts and scholars. They fundamentally don’t understand lower-class life. I’m genuinely from the lower class — though not the absolute bottom since I do have an office job. But I’m not middle class either. I’m somewhere in between. Not quite middle, not quite bottom. But according to my standard of living, I’m definitely lower class. Our existence is really difficult. I need to eat. That’s my primary goal. How can I have enough moral sense to consider all these other complicated things?
Q: What are your plans for the future?
A: Right now, I have the feeling that I’ll never get ahead. No matter how hard I work, nothing will come of it. I’m just a cog in the machine. I’ll never stand out in such an enormous system, and I have no influence on that system. I also don’t have any sense of personal accomplishment. My idol has always been the Founder of the Republic [Sun Yat-sen], but I fear that throughout my entire life I’ll never take even a single step toward becoming like him.
Q: What is it you admire about him?
A: He was bold and daring. He was the one who put an end to more than two millennia of China’s feudal imperial system. Actually, I really don’t like the CCP. My grandfather’s younger brother joined the Kuomintang in 1949. Our family ancestors were wealthy landowners, but because of my grandfather’s brother, our family was treated miserably after 1949. That’s why my grandfather would quietly curse the CCP at home, and it’s also the source of the hatred I’ve inherited.
“No matter how hard I work, nothing will come of it. I’m just a cog in the machine.”
Q: Why did you wish to take the civil service exam?
A: Only a general, I think, can give the order to raise gun barrels an inch higher. If I want to become a general, taking the civil service exam is the only path available. Actually, in high school I secretly vowed never to join the military, the Party, or the government. Back then I even hoped to go to Taiwan and join the National Army, but that was impossible to achieve. As I grew up, I realized how absurd those ideas were, which is why I started thinking about the civil service exam. I feel becoming a civil servant is the only way I can have any influence or change anything.
Q: Will you take the civil service exam again in the future?
A: The civil service exam is too competitive now, and I’m a poor test-taker. Given my abilities, the best I can hope for is to apply for positions in small towns and villages, but these places can’t even pay salaries now. So what would be the point of passing the exam for those positions? So I really wish I could find someone who is able to tell me what I should do next.
This is a translation in cooperation with ChinaFile and China Digital Times of an interview published by the exile outlet Mang Mang. The Chinese original can be found here.
Once known for quality print journalism, Italy’s media industry has suffered several financial strain in recent decades that has in many ways weakened professional values. Traditional reporting has increasingly given way to “infotainment” — a trend pioneered since the 1990s by former prime minister Silvio Berlusconi’s Mediaset empire, the country’s largest broadcaster, which prioritizes entertainment over substantive news. Cash-strapped outlets struggle to maintain journalistic standards, resulting in declining salaries for reporters and cautious approaches to digital innovation and AI integration. Against this backdrop of economic vulnerability, China has been strategically expanding its influence throughout Italy’s weakened media landscape.
Despite having nearly 285,000 Chinese residents, Italy has few Chinese-language media outlets. Meanwhile, collaborations between Chinese state media and Italian news agencies have facilitated the spread of Beijing’s narratives into mainstream discourse. To better understand the complex interplay between Italian media and Chinese state narratives and media engagement, we spoke to Italian journalist Giulia Pompili. As one of the few journalists who has critically covered the on-and-off saga of Italy’s involvement in Xi Jinping’s signature Belt and Road Initiative, Pompili offers a unique perspective on Beijing’s information strategy and its ongoing impact on the Italian media landscape.
Dalia Parete: When we look at media landscapes globally, each country has its unique characteristics and challenges. What are the most important things to understand about how media works in Italy today?
Giulia Pompili: The main challenge for Italian media is financial. Print media do not have the income or budgets they once had when they had large paid-for circulations. Television is still a strong presence in the media landscape. But over the past 40 years, it has increasingly shifted toward “infotainment” — a blend of information and entertainment. This means fewer programs are focused on delivering substantive information, and more segments are designed primarily to entertain viewers rather than inform them.
Regarding the “infotainment” trend in Italian media, [former Prime Minister] Berlusconi pioneered this transformation. He fundamentally changed how Italians consumed information with his three television channels under Mediaset Italia S.P.A. He was also the first in Italy to envision using media manipulation to cultivate public support.
After Berlusconi, all Italian channels, including the national public broadcasting company Rai, considered the “Italian BBC,” transformed the way they presented information to follow the Mediaset path. So, there is now more “infotainment” and less information across the board.
A young Silvio Berlusconi at the Mediaset headquarters. SOURCE: RAI.
Newspapers lost many readers in the early 2000s, and printed information experienced a major crisis at that time. In the past decade, Italy has attempted to expand into digital media through websites and social media. But it has lagged behind countries like the United States. We’ve also seen the rise of influencers and information websites that often translate foreign articles. More recently, informational podcasts have gained some traction, but the business model remains unclear. No one has figured out how to monetize these platforms effectively. Nevertheless, this shift has once again changed the media landscape.
DP: What significant challenges and transformations do you see on the horizon for Italian media? For instance, how are developments like AI or changing consumption patterns affecting the industry?
GP: One of the biggest challenges is declining compensation for journalists. For example, if you are a freelancer, you cannot afford to pay rent for an apartment. And if you are a staff writer or a TV producer, you likely have a very low monthly income.
Italy remains quite conservative in the media sector. AI hasn’t been widely implemented in newsrooms, and significant fear surrounds it. Whenever I discuss this with colleagues, especially those from older generations, they express the concern that AI will take their jobs.
From the consumption side, Italy has a significant information literacy gap because there is no education on media literacy. Most of the population is illiterate when it comes to media. They struggle to distinguish between information from influencers, reporters, staff writers, investigative journalists, and activists. This is especially problematic among younger generations, who often can’t differentiate between a TikTok influencer discussing Uyghurs in China’s Xinjiang region, for example, and a professional journalist who has thoroughly investigated the topic. Ideological perspectives create substantial barriers between activists, influencers, and traditional journalists — representing one of our biggest challenges.
Younger generations often can’t differentiate between a TikTok influencer discussing Uyghurs in China’s Xinjiang region and a professional journalist who has thoroughly investigated the topic.
DP: Despite Italy hosting nearly 285,000 Chinese nationals, few Chinese-language media outlets exist. What factors, in your view, have contributed to this limited media presence, and how does this affect information flow within the Chinese community?
GP: In Italy, it often seems that the large Chinese diaspora is already closely aligned with the Chinese Communist Party — though there is no specific research or data about this. They may not demand dedicatedmediabecause the Party already maintains a strong influence over diaspora groups, and pro-China content is known to dominate the media that are present. The main Chinese-language radio station, China FM Italia focuses primarily on entertainment rather than news. Another outlet, Cina in Italia (世界中国) began as a book publisher. They tried to publish educational books in Italian and Chinese. It was originally a cultural company, but it has now changed its business model, working directly with the official China News Service [under the United Front Work Department of the CCP].
Another unique character of the Chinese community in Italy is that you rarely hear any form of dissenting opinion. As the white-paper protests that began in Shanghai spread internationally in late 2022, there were attempts to organize demonstrations in major squares in Bologna and Rome — but these barely made an impact. Compared to similar protests in Germany, France, and the UK, which were much larger and more visible, the level of dissent in Italy was negligible. In Italy, such activities are notably absent.
DP: So, how would you characterize China’s approach to media or media engagement in Italy?
GP: The media engagement approach has been simple. Embassy personnel have built relationships with Italian editors, editors-in-chief, press agencies, and individual journalists.
Before 2019, Chinese media had numerous bilateral contracts and cooperation programs between Chinese and Italian media. We engaged significantly with the official China Media Group [under the CCP’s Central Propaganda Department], which maintains the most prominent presence everywhere. In reality, they were paying for advertising in Italian newspapers. They would pitch original Chinese-language articles translated into Italian. Like everywhere in Europe, they tried to coordinate with Italian media outlets to publish Chinese dossiers written by the embassy or agencies working with the embassy. Generally, they attempted to use Italian media as a powerful tool to share their narratives.
In 2019, something changed. Xi Jinping came to Italy for an official state visit. During that visit, Italy officially joined the Belt and Road Initiative, establishing numerous institutional cooperation agreements. One of the most notorious agreements for the media was between Agenzia Nazionale Stampa Associata (ANSA), our country’s leading news agency, and the Chinese government’s Xinhua News Agency. ANSA is a primary news source for Italian journalists, so this partnership allowed Chinese state narratives to directly enter Italy’s mainstream news ecosystem.
ANSA’s CEO, Stefano De Alessandri, and former Xinhua’s President, Cai Ming Zhao (蔡名照), signing a cooperation agreement between the two agencies. SOURCE: ANSA.
DP: How did the COVID-19 pandemic affect China’s information strategy in Italy?
GP The COVID-19 pandemic was a turning point for Italy, revealing China’s information manipulation tactics more sharply. A key example was when the Chinese Red Cross sent masks to Italy. Our former Foreign Affairs Minister Luigi Di Maio, who had signed the Belt and Road MOU the previous year, was entirely absorbed by Chinese propaganda and disinformation to rehabilitate China’s image as the country where the virus originated.
We engaged significantly with the official China Media Group, which maintains the most significant presence everywhere.
By late March 2020, China was building its image as Europe’s savior. This tactic worked quite well in Italy. They manipulated the situation by sending masks and supplies to the Italian Red Cross, creating a major political event. However, these weren’t donations but rather purchases made by Italy. We still have numerous legal proceedings regarding emergency funds spent on Chinese supplies. The critical point is that during this emergency, China used Italy as an experiment to see how effectively they could manipulate information to craft their image as a savior amid the pandemic.
An article in Italy’s Il Foglio, published during a visit to the country by Xi Jinping, bears the headline: “We are not in Beijing,” after Chinese diplomats demanded positive coverage.
DP:How did Chinese officials typically engage with foreign journalists, like yourself, who were critical of their policies?
GP: At the time, I was one of the journalists who extensively covered Chinese-Italian bilateral relations. I was also among the few who criticized Italy’s joining the Belt and Road Initiative. At the time, the appointed spokesperson of the Chinese embassy confronted me at the Quirinale Palace during Xi’s visit. He aggressively told me, “You must stop saying bad things about China.” The next day, we published the news headline, “We are not in Beijing.” In the article that chronicled this confrontation, we included the spokesperson’s full name, which made him very angry.
He aggressively told me, “You must stop saying bad things about China.”
This incident also marked the first time that the Italian political establishment realized that the silencing of journalists was something that could not go unanswered.
DP: Despite claims of a tougher stance toward China since Italy’s exit from the Belt and Road Initiative, how would you assess the reality of Italian-Chinese relations, notably regarding media partnerships and Meloni’s broader political agenda?
A page at Italy’s Agenzia Nova dedicated to coverage by China’s official Xinhua News Agency.
GP: We are saying that we are restricting Chinese influence, right? Italian printed media generally reduced Chinese content partnerships, but some outlets still publish Chinese state-sponsored content for financial compensation. While the “Chinese dossiers” appear less frequently, Italy remains an outlier in Europe by continuing to monetize the publication of Chinese government messaging in its media landscape.
The Chinese government’s official Xinhua News Agency changed cooperation partners from ANSA to Agenzia Nova, a popular online news source. So, it is still doing what it was doing with new partners.
From a political perspective, Meloni’s core focus as Italian president is immigration — she doesn’t think about much of anything else. She knows that China is the only country that can help her in Africa because China currently has the most significant political influence there.
She understands that she cannot effectively deal with Libya, Algeria, or Egypt without support from Chinese officials and institutions. For Meloni, the only priority is this very concrete issue, and she is ready to do whatever it takes to achieve her singular foreign policy goal: managing immigration. She knows that she needs China to stabilize the relationship with Africa.
In a move that highlights China’s expanding use of institutions down to the local level to supercharge the state’s global influence efforts, the School of Foreign Languages at Xi’an Jiaotong University, a leading public university in the central Chinese city, has formed a partnership with a center run by the province’s propaganda office.
On March 5, the Shaanxi International Communication Center (陕西国际传播中心), or SICC, a provincial-level external propaganda office formed in December 2023 through the state-run Shaanxi Radio and Television Group — directly under the propaganda office of the provincial CCP committee (中共陕西省委宣传部) — announced that it would partner with the university to work on international communication issues using “interdisciplinary approaches.” The goal, according to state media coverage, is to provide “talent guarantees” for “telling Shaanxi stories” and promoting Chinese culture internationally, both references to the broader the state-driven goal of expanding China’s global influence, which Xi Jinping has advanced under the phrase “telling China’s story well” (讲好中国故事).
School of Foreign Languages Director Chen Xiangjing (陈向京) and SICC chief Wang Dong (王冬) sign the agreement for cooperation on external propaganda.
The agreement, signed by School Dean Chen Xiangjing (陈向京) and SICC Director Wang Dong (王冬), who is also a ranking provincial propaganda official, further underscores the serious ethical issues at play as China blurs the lines between higher education and state propaganda in its efforts to address a shortage of talent for its ambitious external communication goals as set by Xi Jinping.
Since its formation in late 2023, the SICC claims to have built a “global communication matrix” targeting neighboring countries in Central Asia through branded platforms on social media sites such as Facebook and YouTube, both services that are blocked in China.
The SICC also, like a number of China’s expanding international communication centers, has a vertical relationship with China Daily, the state-run newspaper directly under the State Council Information Office (in practice, the very same office as the Central Propaganda Department) whose primary role is to advance the government’s external communication.
According to the agreement signed last week, the two sides pledged to establish a “dual-purpose base” (双基地) for talent cultivation and research, including a “Graduate Collaborative Education Base” (研究生协同培养育人基地) between the university and Shaanxi Broadcasting and Media Group, as well as the launch of an “International Communication and External Discourse Innovation Center” (国际传播与对外话语创新中心).
Effusive remarks about spicy chicken and boiled fish, and about a light rail train threading right through the center of an urban apartment building — these may not at first seem like the makings of a strategic state-led soft power push. But they are central to the larger narrative of how China is seeking to influence hearts and minds in Indonesia, and through the rest of Southeast Asia, by mobilizing local social media influencers.
Last month, a group of 10 Indonesian influencers, each with audiences topping hundreds of thousands of followers, toured China’s inland municipality of Chongqing, marveling at its dazzling skyline, vibrant nightlife, and cultural heritage. Their social media feeds soon filled with breathtaking images and glowing reviews of the city.
These influencer trips, organized from the central level down to a new generation of state-run local “international communication centers,” are part of a sophisticated Chinese soft power strategy to reshape its image in Indonesia, where historical skepticism toward China persists despite growing economic ties. By leveraging the voices of popular Indonesian figures with massive online followings, China aims to complement traditional diplomatic channels and create a more nuanced narrative that portrays China as a modern, culturally rich partner while potentially shifting attention away from more complicated questions about its regional influence.
Roots of Unease
Indonesia has long grappled with deep-rooted anti-China sentiments, a legacy of historical tensions, political narratives, and economic competition. In recent years, as Chinese investments have surged in the archipelago — funding everything from infrastructure to digital technology — skepticism toward China has grown. Many Indonesians worry about economic dependency, job displacement, and geopolitical influence. Against this backdrop, China’s effort to shape its image in Indonesia is not an incidental occurrence; it is a calculated strategy in which Indonesian influencers now play a crucial role.
Over the years, China has carefully maintained its positive narratives in Indonesia, centering its efforts on two key objectives. First, it seeks to present its Belt and Road Initiative (BRI) projects and economic activities in Indonesia as overwhelmingly beneficial. Late last year, for example, China called the cooperative Jakarta-Bandung High-Speed Rail a “golden signboard of China and Indonesia’s joint Belt and Road construction,” and suggested that it was “an important engine” driving Indonesia’s economic development. In addition, Xinhua’s Indonesian feeds on the social media platform X often contain translated Xi Jinping’s speeches and narratives on how the BRI would benefit Indonesia. The full picture of the project is far more complicated, with lingering concerns about delays, cost overruns and state debt.
Generally, China’s state narrative of mutual benefits for Indonesia highlights job creation, infrastructure development, and economic growth, positioning China as an indispensable partner in the country’s progress. Second, given Indonesia’s status as the world’s largest Muslim-majority nation and the growing criticism of China’s policies in Xinjiang, Beijing aims to deflect scrutiny and reshape perceptions regarding its treatment of the country’s Uyghur Muslim minority.
To achieve these objectives, China has actively invited not only social media influencers but also journalists, Muslim scholars, students and academics on carefully curated trips to China. These initiatives aim to shape their perspectives and, in turn, influence public discourse in Indonesia by promoting a favorable image of China. Visitors return with firsthand experiences, often highlighting China’s technological advancements, cultural richness, and religious freedom, while avoiding politically sensitive topics that could complicate the narrative.
The strategy is particularly targeted when it comes to religious concerns. When inviting Muslim scholars, in particular, China has placed a strong emphasis on Xinjiang, often showcasing what it claims to be the peaceful coexistence of Islam within the country. These trips feature visits to state-approved mosques, meetings with Chinese Muslim communities, and narratives that seek to downplay repression in the region, which has been well-documented. The goal is to reassure Indonesian audiences, particularly its large Muslim population, that reports of human rights abuses in Xinjiang are exaggerated or misleading.
A Case Study in Influencer Diplomacy
The recent visit by a group of Indonesian content creators to Chongqing seems at first glance to be a feel-good tour focusing on local cuisine, modern wonders and traditional culture. But the visit exemplifies the influencer strategy in action, and shows how increasingly local and regional propaganda offices from China are also becoming involved in such efforts across Southeast Asia.
The group explored the city’s towering skyscrapers, neon-lit skyline, and historical sites, observing how modern architecture blends with traditional elements.
At the Hongyadong Scenic Area, overlooking the Yangtze River, the influencers took in the city’s illuminated waterfront. Reflecting on the visit, Maria Asteria Sastrayu Rahajeng, an Indonesian beauty pageant titleholder, noted Chongqing’s scale and infrastructure, while her sister, author, news host and influencer Elizabeth Krisansia S. Rahajeng described the city’s nighttime views and lively atmosphere the most effusive terms.
Indonesian model and influencer Maria Rahajeng poses at Chongqing’s Cloud Eye Observation Deck on February 15, 2025. SOURCE: Instagram.
The itinerary included a guided walk through key landmarks like the Liziba Monorail, which appears to pass through a residential building, and The Ring Shopping Park, a commercial space incorporating urban greenery. At the Sichuan Fine Art Institute, they explored street murals and learned about Chinese tea culture. Kyra Nayda, a full-time digital content creator based in Jakarta, remarked on the city’s mix of modern technology and traditional aesthetics, and promoted heavily to her online audience the stilt-building complex known as Hongya Cave (洪崖洞), a popular tourist destination whose towering teahouse feel has drawn comparisons to the grand bathhouse in the Japanese animated film Spirited Away. “It gives off cyberpunk vibes,” Nayda posted to her fans, “like a city from a sci-fi movie.”
Throughout their journey, the influencers continuously took photos and uploaded content to their social media platforms, sharing their experiences in real-time and engaging with their followers. With each post, they served as cultural ambassadors carrying China’s message back to Indonesia’s digital landscape.
For Cindy Karmoko, who identifies herself as a digital creator and “fragrance enthusiast,” the visit highlighted aspects of China she had not previously encountered. She noted the city’s size, transportation system, and climate, as well as its culinary diversity beyond its well-known hot pot. The trip also included visits to Chongqing’s mountainous areas and experiences such as a cableway ride across the Yangtze River and a walk through the city’s historic 18-step alleys.
The tour was arranged, according to promotional materials posted to social media, by Jakarta-based Indah Tour. But a post by the tour company made clear that it was directly supported by the Chinese embassy.
A poster for the February tour of China by Indonesia influencers, organized by Indah Tour, but clearly showing Chinese embassy support. SOURCE:Facebook.
You might ask: Is it not standard practice for government’s to have a role in tourism promotion? Consider that eight years ago, Indah Tour sent a similar group of influencers, including Elizabeth Rahajeng, to Taipei to promote tourism to Taiwan under a “Time for Taiwan” campaign sponsored by the Taiwan Tourism Bureau under the country’s Ministry of Transportation and Communications. Are these not recognizable tourism promotion tactics?
But the clear difference in China’s case is how closely such events are tied to state-led promotional efforts at both the national and local level. On February 18, as the tour was underway, China’s official Xinhua News Agency released a report on the trip, accompanied by a video that also went out through Xinhua’s GlobaLink service. The reports were released to Indonesian media through Antara, the national news agency, which since 2016 has had a close partnership with Xinhua. In Chongqing, meanwhile, the tour was coordinated and covered by the Western China International Communication Organization (WCICO), a center run under the propaganda office of Chongqing’s CCP leadership. Also known as the Chongqing International Communication Center (重庆国际传播中心), WCICO is part of a growing network of local and regional offices now tasked with pushing China’s state narratives through connections at a more grassroots level.
Soft Messages
The response from Indonesian audiences to these influencer-driven narratives appears largely uncritical, at least in public discourse. Posts from participating influencers, such as Rahajeng’s reflection on “China’s power, strength, and resilience,” have received positive engagement, but mostly in the form of generic likes, heart emojis, and admiration for the aesthetics of the trip. There is little visible debate or discussion about the broader implications of China’s growing influence in Indonesia. This suggests that while influencer narratives are reaching wide audiences, they may not be prompting deeper reflection on the geopolitical dimensions of China-Indonesia relations.
At the same time, mainstream Indonesian media coverage of these trips closely echoes Chinese state narratives, offering little scrutiny. Reports from national news agency Antara, which has the longstanding Xinhua partnership, highlight the beauty and modernity of destinations like Chongqing and Zhangjiajie but omit any discussion of China’s broader strategic interests. This alignment between influencer content and media narratives suggests a highly curated information environment — one where China’s soft power messages are disseminated effectively but with minimal critical engagement from Indonesian audiences.
China has said it pursues a level of “national cultural soft power” (文化软实力) commensurate with its perceived role as a rising global power. This reference to “soft power,” which the leadership has used internally since 2007, echoes the concept popularized by political scientist Joseph Nye, referring to a country’s ability to influence others through attraction rather than coercion. But far from encouraging the more natural and grassroots-driven exchange of ideas that Nye’s concept entails, Beijing has taken a managed approach — preferring state-driven economic partnerships and cultural diplomacy over real choices that give agency to the local populations potentially impacted by China’s growing presence.
The stilt-building complex known as Hongya Cave (洪崖洞), a popular tourist destination in Chongqing. Source: Instagram.
China often defines its “soft power” project against the immense influence of the US-led West, bemoaning the outsized influence of Western media. But China’s hard-nosed — even “sharp” — approaches to its soft power initiatives, put developing countries like Indonesia in a difficult position, particularly given their own challenges when it comes to media health and information integrity. Many of China’s actions, including the frequent diplomatic visits to local Indonesian media made by the country’s foreign ministry and stressing concepts like “shared destiny” and the need to “tell China’s story well,” are less about public diplomacy and open exchange, and more about loosely coercive influence.
The recent influencer trip to Chongqing is another textbook example. By leveraging the reach of young, popular Indonesians, Chinese state actors at the central and regional level hope to cultivate a new generation of Indonesians who see the country not as a looming economic threat, but as a land of opportunity, beauty, and modernity.
The rise of Indonesian social media influencers has given China a unique opportunity to shape perceptions in a way that feels organic. Unlike traditional government-led PR campaigns, influencer marketing capitalizes on trust. Indonesians are among the most active social media users in the world, and influencers are often seen as more relatable and credible than traditional media or official statements. Their firsthand experiences in China, shared through vlogs, Instagram stories, and TikTok videos, have the power to normalize positive narratives about China among their millions of followers.
The 2024 China-Indonesia Media Forum focused on media ties as a bilateral state-level issue, involving major state-run Chinese media and proposing a community of influencers to promote ties.
Far from being an improvised tactic, the idea of engaging with social media influencers is rooted in the discussions at last year’s China-Indonesia Media Forum, where the role of influencers in fostering closer ties between the two nations was explicitly addressed. During the forum, the proposal to form an Indonesia-China influencer community was introduced as a way to further strengthen bilateral relations. This acknowledgment underscores that China’s engagement with Indonesian influencers is not a spontaneous trend but rather a carefully planned strategy within a broader diplomatic framework.
This influencer-driven approach allows China to sidestep traditional diplomatic and media channels, which are often met with skepticism. Instead, audiences see authentic-seeming testimonials from personalities they admire, making the messaging more persuasive. For younger Indonesians, who consume news and cultural content primarily through social media, these influencers become key opinion leaders shaping how China is perceived.
Challenges and Limitations
However, this strategy is not without risks. While these curated trips paint a picturesque image of China, they do not address the real concerns of many Indonesians—ranging from the treatment of Uyghur Muslims to the ongoing South China Sea tensions. Furthermore, social media endorsements, no matter how visually compelling, can be met with skepticism, especially when audiences suspect a promotional agenda behind them.
Transparency is key to maintaining credibility. If Indonesian influencers fail to disclose the nature of their trips, their audiences may begin to question their motives. The backlash against influencers who are perceived as too promotional or politically motivated can be swift and unforgiving. As China continues to engage Indonesian influencers as part of its soft power agenda, both parties will have to navigate this delicate balance carefully.
Another challenge lies in the broader landscape of digital influence. Many Indonesian influencers have built their brands on authenticity and relatability. If their association with China appears overly orchestrated or politically motivated, they risk alienating their audiences. This could ultimately undermine the very soft power strategy China hopes to deploy, reinforcing existing distrust rather than alleviating it.
Indonesia finds itself at a crossroads. As Chinese investments continue to shape the nation’s economy, the battle over public perception intensifies. Will China’s assertive soft power efforts succeed in rewriting its image in Indonesia through social media influencers? Or will the deep-seated distrust prove too formidable a challenge? One thing is certain: Indonesian influencers are now at the center of this influence game, and in a digital age where narratives are shaped in the palm of our hands, their role will only grow in significance.
As Beijing and Jakarta navigate their complex relationship in the coming years, perhaps the most powerful diplomatic tools won’t be found in formal state meetings or economic agreements, but in the carefully filtered Instagram posts and TikTok videos of Indonesia’s digital stars. In this new era of influencer diplomacy, the boundaries between soft power, marketing, and international relations have never been more blurred — or more consequential.
