Chinese Propaganda Operations Caught Using AI to Target Taiwan, Revealing Strategic Failures

The Chinese Communist Party’s (CCP) efforts to manipulate public opinion through social media have taken an embarrassing turn, as researchers discovered AI-generated propaganda posts targeting Taiwan with the prompts still visible in the published content.

Wang Hung-en, a researcher at RAND Corporation and associate professor at the University of Nevada, Las Vegas, recently uncovered several Facebook pages publishing AI-generated content in traditional Chinese specifically designed to influence Taiwanese audiences. In a striking oversight, operators forgot to remove the AI prompts embedded in the final text.

“One striking example came from a page introducing Chinese public figures,” Wang noted in a Facebook post. The article concluded with an unedited instruction: “This article has been re-edited for Taiwanese users, using Traditional Chinese characters, with the word count limited to 500. The beginning of the original text has been retained, and the historical authenticity of the original text has not been altered.”

The discovery confirms what cybersecurity experts have long suspected: Beijing is increasingly using artificial intelligence to scale its influence operations against Taiwan. Rather than targeting broad overseas Chinese audiences, these operations specifically customize content for Taiwanese readers.

Wang’s analysis shows many of the articles are saturated with Chinese nationalism and anti-Japanese messaging, closely aligned with Beijing’s official propaganda. The content appears connected to the Cyberspace Administration of China and the CCP’s propaganda apparatus. The use of AI helps conceal the origin of posts, making attribution more difficult.

The revelation sparked significant engagement online, with many netizens saying they had noticed suspiciously similar writing styles proliferating across social media. “Recently I’ve seen so many posts like this circulating on social media. I always thought it was suspicious — now the truth is finally out,” wrote one commenter. Another added: “I’ve seen plenty of them here in the U.S., and they really do influence people.”

This discovery comes as Taiwan’s National Security Bureau (NSB) recently completed a report titled “Analysis of CCP Cognitive Warfare Operations Against Taiwan in 2025,” outlining five major operational methods employed by Beijing.

According to the NSB, CCP party-state-military systems first establish controversial narratives, then mobilize technology firms, PR companies, and cyber groups to spread targeted messaging. Agencies including the Cyberspace Administration of China, the Ministry of State Security, and the PLA Political Work Department employ firms such as “Zhongke Tianji” and “Meiya Pico” to collect data on Taiwanese politicians, opinion leaders, and election trends.

The NSB report further details how the CCP creates fake international media sites to amplify Beijing’s narratives, cultivates content farms on Facebook, and deploys “soft topic” accounts on platforms like Threads and X to subtly inject political messaging.

Perhaps most concerning is the Ministry of Public Security’s use of the “dragon bridge” cyber army group, which conducts influence operations in more than 20 languages across 180 global platforms. This effort is supported by automated systems managing tens of thousands of fake accounts.

The NSB also warns that Chinese state-linked enterprises are accelerating AI development to integrate public opinion monitoring, automated video generation, and precision targeting. Companies such as iFlytek have reportedly been tasked with developing voice systems to lure Taiwanese citizens into recording samples, later used for synthetic impersonation.

Taiwan’s government has not been passive in response to these threats. The NSB noted it has held more than 80 security dialogues and special meetings with international partners to strengthen democratic cooperation against cognitive warfare, while continuing efforts to expose and remove disinformation.

The revelation of poorly executed AI propaganda highlights a key vulnerability in China’s information operations. Despite substantial resources invested in these campaigns, basic operational errors suggest the propaganda apparatus may be struggling with quality control as it scales up its digital influence operations.

As AI technology becomes more sophisticated, detecting such influence operations will likely become more challenging, underscoring the importance of international cooperation in identifying and countering state-sponsored disinformation campaigns targeting democratic societies.