This report is based on a joint study by the Civil Network OPORA and the Taiwanese organization Factlink. You can read OPORA's study on the use of AI to promote Russian propaganda on TikTok, created as part of this collaboration, at the link. Factlink's study, also produced within this collaboration, can be read at this link.

Authors: 

Olha Snopok, Analyst, Civil Network OPORA

Olha Bedovska, Analyst, Civil Network OPORA

Wei-Ping Li, Research Director, FactLink

Mary Ma, Head of Investigation, FactLink

Summer Chen, Founder and reporter, FactLink

1. Introduction

What has generative AI changed in Russia’s information operations against Ukraine and in the Chinese Communist Party (CCP)’s information operations targeting Taiwan on TikTok? This report answers this question by bringing together two studies that approach it from different angles. The Ukrainian case shows how AI-driven propaganda works in a society at war; the Taiwanese case, in a society living under a persistent threat of war.

Ukraine has been confronting Russian aggression since 2014, when Russia annexed Crimea and started the war in Donbas. In February 2022, this war escalated into a full-scale invasion. Since then, Ukraine has been fighting along a long frontline, parts of its territory remain under occupation, and cities across the country face regular missile and drone strikes, including on energy infrastructure. The war has affected every part of society. Millions of people have been displaced within the country or abroad, and the army depends on ongoing mobilization. Political life has changed as well. Ukraine has been under martial law since 2022, and under Ukrainian law, no elections can be held until it is lifted. Conscription, the course of the war, and the prospects for peace have become central issues in public debate, and they are also among the issues most exposed to outside manipulation.

Information warfare has been part of Russia’s aggression from the start. Russia uses a wide range of channels, from state television and fake news websites to Telegram channels and networks of social media accounts. Through them, it seeks to undermine trust in Ukrainian institutions and the armed forces, deepen social divisions, and weaken international support for Ukraine. Ukraine has built a response that brings together state bodies, independent media, fact-checkers, and civil society organizations, but the scale and adaptability of Russian operations remain a constant challenge.

In Taiwan’s case, after the CCP has ruled China for 77 years, the CCP has stepped up its efforts to achieve the goal of “unifying” Taiwan in recent years. Alongside intensified military drills around Taiwan and frequent, deliberate micro-aggressions by fighter jets and drones crossing the Taiwan Strait median line, the CCP has waged information interference campaigns against Taiwan. According to reports by V-Dem, a University of Gothenburg program that assesses democracies, Taiwan is the country most targeted by disinformation campaigns waged by foreign states. The CCP’s information interference tactics include not only blatant propaganda portraying China as strong and rich while taunting Taiwan as weak and fragile, but also insidious disinformation campaigns aimed at deepening divisions within Taiwanese society. Over the past two years, evidence has shown that malicious actors have also used AI to create fake accounts and spread disinformation during controversial incidents, and even to try to sway public opinion during elections.

Despite these different contexts, both countries face the same shift. Generative AI has made disinformation faster, cheaper, and easier to produce at scale, and it is already shaping the intensity of influence operations in Ukraine and Taiwan alike. 

This report draws on two studies. Civil Network OPORA’s study, What Role Does AI Play in Russia's War Against Ukraine?, analyzed 240 TikTok accounts that used AI to produce content on four themes, from undermining trust in the military and the authorities to religious and seemingly pro-Ukrainian content. FactLink’s study, A Bridge Built in the Clouds: AI-Generated Propaganda Promoting Taiwan–China Reunification on TikTok, examined AI-generated videos that present the Xiamen–Kinmen Bridge as a symbol of Taiwan’s “return” to China. By bringing these findings together, the report aims to identify the similarities and differences in how Russian and Chinese propaganda use AI to influence audiences in Ukraine and Taiwan.

2. Methodology and Limits of Comparability

This report brings together two studies that were designed and conducted independently. OPORA examined AI-generated content across four thematic areas in the Ukrainian segment of TikTok, while FactLink focused on a single theme, the Xiamen–Kinmen Bridge, in content aimed at Taiwanese audiences. The two studies therefore differ in scope, sampling, and the way AI-generated content was identified. This section describes each approach in turn and then outlines the limits these differences place on comparison. For this reason, the report compares tactics and narratives rather than the scale or reach of the two campaigns.

OPORA  report’s research methods

As TikTok provides researchers with very limited tools for content analysis, identifying and tracking actors, their behavior on the platform, and the narratives they share is difficult. Given these limitations, OPORA began with exploratory research on several selected topics. Initially, the study focused on two areas: undermining the reputation of the Ukrainian military and undermining the reputation of the Ukrainian authorities. For each area, OPORA compiled a list of hashtags that, in the team’s view, reflected most of the public discussion on these topics. While browsing the desktop version of TikTok, however, the team noticed that the “You may like” section systematically recommended profiles with AI-generated content on two further themes: pro-Ukrainian content and content promoting narratives associated with the Russian Orthodox Church. Because the platform’s algorithm linked these themes to the original ones, OPORA broadened the study to include them.

Using the Apify service, OPORA downloaded videos containing the selected hashtags. The team then reviewed these videos manually to identify those created with the help of AI, as well as the profiles that systematically posted such content on one of the selected topics. Profiles for the two additional areas were identified through TikTok’s recommendations. The identified profiles were added to a separate database for more detailed analysis.

OPORA classified a video as AI-generated only when this was unmistakable from its content, for example, in animations or in scenes depicting events that never happened. The team relied on visual assessment rather than automated detection tools, and videos whose origin remained uncertain were not counted as AI-generated. Because of this conservative approach, the study is more likely to underestimate than to overstate the use of AI.

OPORA collected 60 profiles per area (240 in total) and downloaded all 152,759 posts these accounts had published from their creation through July 13, 2026. From this dataset, the team drew a sample of 2,000 posts per area, stratified by account so that every profile was represented proportionally. This sample captures the key tactics used in AI-generated content, but it does not show how the use of AI changed over the observation period. Some accounts were deleted during the study, so the final figures may vary. Because TikTok does not show when an account was created, OPORA used the date of each account’s first post as a proxy. This approach is approximate: some accounts may have been created earlier but remained inactive, while others may have deleted older posts. 

Finally, OPORA manually analyzed the resulting dataset of 8,000 posts to identify the key themes and narratives that various actors use to influence Ukrainian TikTok users in the context of Russia’s war against Ukraine.

FactLink report’s research methods

FactLink selected cases illustrating a typical Chinese AI-powered information manipulation campaign. The team focused on AI-generated TikTok videos about the CCP’s ongoing and controversial plan to build the Xiamen-Kinmen Bridge, which the CCP plans to physically connect Xiamen, a city in Chinese territory, with Kinmen, an island governed by Taiwan. The Taiwanese government views the bridge as a threat to its national security because it could facilitate a CCP invasion. Taiwan has since refused to participate in the construction. However, the CCP has built a bridge connecting two islands in Xiamen and claims it will continue the work to extend it to Kinmen. Alongside the construction, AI-generated propaganda videos have portrayed the bridge as evidence of a shared blood connection and a symbol of “the road for Taiwanese to go home [China].” The videos also claim the bridge will give Taiwanese people, particularly Kinmen residents, convenient access to better resources in China and greater economic prosperity. The FactLink team analyzed the videos' narratives and how their producers used AI to promote propaganda, including inaccurate information, to influence viewers.  

To conduct this research, FactLink searched Apify and TikTok for relevant videos posted between 2023, when the Chinese government began building the Chinese side of the bridge, and September 1, 2026, using the keywords “Xiamen-Kinmen Bridge” and “Kinmen-Xiamen Bridge.” The search yielded 112 videos focused on the topic. FactLink team then manually reviewed and labeled each video by account location, posting date, audience engagement, subtheme, video style, and whether it was AI-generated. To identify trends and possible coordination among creators, the team members compared video posting dates, account locations and creation dates, and similarities in images and scripts, then extracted shared elements to identify common narratives.

Parameter

OPORA

FactLink

Focus

4 thematic areas (the army, the authorities, the Russian Orthodox Church, pro-Ukrainian content)

1 topic: the Xiamen–Kinmen Bridge

Volume

240 accounts, 110k+ posts, 8,000 analyzed manually.

112 videos, 23 of them part of a coordinated campaign.

Period

2021 – July 13, 2026

2023 – September 1, 2026

Search method

Hashtags and keywords via Apify and TikTok + “You may like” recommendations

Keywords via Apify and TikTok

Independent AI verification

Not applied

AI detectors, assessment conducted by FactLink team members and external experts

Account age

Date of first post as a proxy

Account creation dates

Coordination

Not tracked

Inferred from scripts, dates, and locations

3. Shared Tactics

Both OPORA and FactLink studies document five tactics found in the Russian and Chinese cases alike.  Russian and Chinese propaganda rely on very similar tactics but fill them with different narratives, which we discuss further in the next section. 

In both cases, propagandists aim to mislead the audience by mixing AI-generated content with authentic content. Additionally, both studies found signs that in some cases AI-generated social media content may be amplified through coordinated rather than organic activity. To influence TikTok users, they employ fake hosts presented as credible authorities to make their narratives appear more trustworthy and convincing. They use generic, lighthearted, or humorous content to attract apolitical audiences, while sensitive and emotionally charged content is designed to target specific identity markers. Let's look at each tactic more closely. 

Tactic 1. Mixing AI-generated content with authentic content

Both OPORA and FactLink found that AI-generated content is often published alongside real footage. In the long term, this is the most dangerous technique, as it blurs the line between what happened and what did not.

In Ukraine, a significant amount of AI-generated content is spread by accounts that mix generated videos with real ones on their pages. This has been particularly prominent in TikTok content focusing on mobilisation. The same creators posted real videos of citizens being detained by Territorial Recruitment Centers (TRCs), followed by AI-generated ones on similar or closely related topics. Also, the real humorous scenes involving TRCs were mixed with mocking AI-generated songs and videos. In one TikTok feed, you can see real and generated videos in a row, with real of them claiming that “the army raises slaves, not men”, and another using generated heartbreaking songs about veterans. A similar pattern can be observed in videos featuring veterans’ and soldiers’ real traumatic experiences, which were mixed with fabricated ones. Some TikTok accounts belonging to school pupils or lesser-known bloggers try to gain more followers by deliberately posting provocative AI-generated content. They likely do this because they understand that engagement-bait content makes their profiles more visible and helps them appear more often in users' feeds.

A similar pattern appears in the Taiwanese case.  Since 2025, AI-generated scenes of the  Xiamen–Kinmen Bridge construction have been added and mixed with authentic images. As an illustration, FactLink noted AI-generated scenes with Chinese builders working in harsh weather conditions. 

As both organizations point out, combining authentic and fabricated content blurs the line between what is real and what is fake. When synthetic videos appear alongside genuine material, viewers find it increasingly difficult to tell them apart. In some cases, this may lend fabricated stories a credibility they would not have on their own. In others, it may leave viewers with the sense that “things are not so clear-cut,” leading them to doubt genuine information as well.

Tactic 2. AI hosts as a way to appeal to authority

Both OPORA and FactLink documented that Russian and Chinese propaganda speak through presenters who do not exist. By mimicking authoritative figures and familiar news formats, such content makes disinformation appear more credible.

As OPORA states, imitating news content is a fairly popular technique. Two key approaches were used. In the first, a voiceover delivered the message while the video consisted of AI-generated material, sometimes mixed with fragments of real footage. This also connects to the first tactic described above: mixing real and synthetic content makes the distinction between the two even less obvious. In the second approach, AI was used to create a speaker delivering the news. Such messages were often emotionally charged and contained disinformation.

On the other side of the globe, Chinese propaganda creates TikTok accounts for AI-generated female hosts who claim to be experts in finance and investment. These attractive synthetic women give financial advice, provide market analysis, and post similar videos about the Xiamen–Kinmen Bridge. They read almost identical scripts and share the same visual style.

Beyond making disinformation look more credible, this tactic can also help it reach audiences that would not normally engage with political content. A fake financial expert can reach people interested in investment, while an attractive influencer-like host may reach a completely different audience. In this way, propaganda does not necessarily have to look like propaganda — it can enter people’s feeds disguised as the kind of content they already consume.

Tactic 3. Political messages in non-political packaging

In both cases, propaganda takes different forms to feel closer, more attractive, and easier to digest for wider audiences. We have already described the Taiwanese case above: AI-generated TikTok videos embed pro-Chinese messages in investment tips and financial commentary. 

In Ukraine, OPORA observed a mixing of entertainment and discrediting content. Such messages were often embedded in humorous videos that mocked, demeaned, or belittled politicians or military personnel. Perhaps the most telling example is accounts specializing in AI-generated music. These songs frequently carried narratives portraying Ukrainian politicians as corrupt or irresponsible, Ukrainian servicemen as victims of circumstance, and staff of Territorial Recruitment Centers (TRCs) as pure evil.

In different ways, these formats touch on sensitive and painful topics to make viewers feel strong emotions — from despair to hate or fear. Strong emotions drive people to react: comment, share, or otherwise engage with the content. The goal ofAI-generated videos is not always to change viewers’ minds or convince them directly. Sometimes, getting a reaction is enough: engagement helps the content spread further and puts the same message in front of more users.

Tactic 4. Using AI to scale up narratives

The fifth tactic is also the most technical one. Both organizations observed the most noticeable spikes in 2025–2026: OPORA in the number of new accounts posting AI-generated content, and FactLink in the number of AI-generated videos. One possible explanation for this timing is the rapid improvement and growing accessibility of video-generation models. Producing such video content at scale previously required considerably more resources, while generative AI has substantially lowered these production barriers. This makes it possible to produce large volumes of content around the same narratives much more easily.

However, production is only one part of the process. The reports also point to patterns suggesting that this content may be distributed through coordinated networks of accounts rather than spreading entirely organically.OPORA’s data show that new accounts appeared in waves rather than evenly over time: “in July and September 2025, the number of new channels exceeded 8 and 10, respectively, for the first time, and in March 2026 an absolute maximum for the entire observation period was recorded—13 new accounts in a single month.” In OPORA’s view, “this dynamic may indicate not organic growth of interest in the topic but a situation in which new channels appear in groups during certain periods.”

FactLink’s evidence goes further, because it rests not only on timing but on the content itself. The researchers identified 12 “female expert” personas who posted 23 videos about the Xiamen–Kinmen Bridge using almost identical scripts and the same visual style. Four of these personas appeared on several accounts under different names. One host, for example, posted the same content from three differently named accounts. The timing was just as telling: 21 of the 23 videos were posted in late July and August 2026, and most of the accounts had been created after late 2025. Most of the accounts also appeared to be based in Southeast Asian countries such as Malaysia, Singapore, and Myanmar.

4. Shared Narratives

OPORA and FactLink identified several similar narratives used by Russian and Chinese propaganda in the AI-facilitated TikTok videos analyzed:

Shared narrative 1. “Your government is heartless. – see how good we are”

In comparing AI-facilitated TikTok propaganda videos produced by Russia and those targeting Taiwanese audiences, one of the shared narratives emerges, which portrays the Taiwanese and Ukrainian governments as corrupt, heartless, and hostile toward their people, while presenting the Russian and Chinese governments as strong and caring.

Among TikTok videos produced by Russia, several minor narratives contribute to one theme: portraying Ukrainian authorities negatively. For example, Ukrainian president Volodymyr Zelensky is depicted as a maniac who prevents Ukrainian men from leaving the country, while his children do not fight on the battlefield. In separate videos, Zelensky is shown eating black caviar and drinking wine in the Maldives with money stolen from Germany's military aid. Moreover, some AI videos allege that Zelensky "dragged Ukraine into the war himself,” leaving "the whole country ... burning, with no electricity or water.” In contrast, Russian president Vladimir Putin appears in AI videos as a strong leader who stands firmly with his people and has historically come to Europe as a “liberator.”

The narrative that “the Taiwanese government didn’t treat its people well, but China cares about the Taiwanese people” also appears in AI videos targeting Taiwanese audiences. The videos claim that the Chinese government and ordinary Chinese people have Taiwanese people in mind and are doing their best to “help the Taiwanese go home [note: In these videos, “home” refers to China].” In one video, AI characters used inaccurate information, claiming that China had supplied Kinmen residents with electricity from the mainland to show how much China cares about Taiwanese people. The video also claims that China would do everything it could to complete the bridge because Taiwanese people want a better life, which, according to the video, is what the Taiwanese government has prevented its citizens from pursuing.

 Shared narrative 2. Western partners as unreliable actors

While smearing the Ukrainian and Taiwanese governments, the videos also share another similar narrative to vilify Western allies: Russian propaganda stresses that the US and the EU will abandon Ukraine at the first opportunity because they have grown tired of it; Chinese propaganda portrays the US as a helpless bystander.

For example, some Russian AI propaganda videos claim that Zelensky had “become tiresome” to Donald Trump and was gradually losing U.S. support, while European partners laughed at him and refused to supply weapons. In the Chinese videos examined by FactLink, the US is portrayed as panicked and helpless as it watches China make progress toward unifying with Taiwan. Several videos featured construction work on the “Xiamen-Kinmen Bridge,” with a host or voice-over exclaiming, “The U.S. really can’t sit still anymore! They are in a state of frantic agitation, as Taiwan is on the verge of having a direct link with the mainland [China]”! In one AI-generated scene that appeared across several videos, Trump was surrounded by supporters and praying with them.

Shared narrative 3. Denying Ukraine’s and Taiwan’s agency

Among these videos that promote narratives of a strong Russia or prosperous China and portray Western allies as betrayers, there is a clear denial of Ukraine’s and Taiwan’s agency, as well as a lack of understanding of how Ukrainians and Taiwanese think. Russian propaganda videos one-sidedly stress that Ukraine could not resist on its own and has to capitulate, while Chinese propaganda insists that integration with Taiwan would happen regardless of Taiwan’s unwillingness.

In the Russian propaganda examples, the videos promote the claim that Western partners are gradually turning away from Ukraine and halting weapons supplies, suggesting that defeating Russia is impossible. Other videos make direct, fairly emotional calls to end the fighting, reinforced by AI-generated scenes depicting Russian strikes on Ukraine. In Taiwan’s case, some videos claim, “whether Taiwan wants it or not, we will make the bridge to Taiwan!”—a statement that shows how little the CCP considers Taiwanese people’s views.   

Shared narrative 4. Appeals to a shared identity or a “blood connection”

Interestingly, Russia and the CCP also invoke claims such as “We share the common identity” or “We are connected by our blood” to target Ukrainians and Taiwanese. Russia seeks to construct a narrative in which Ukrainians are merely “misguided” Russians. It also actively targets the Orthodox Church and the Ukrainian language—both central to Ukrainian identity—denying the existence of a Ukrainian church and language and offering Russian alternatives instead. Chinese propaganda, similarly, invokes the “blood of both sides of the Strait” and the “road home.” For instance, the bridge has become a symbol connecting Taiwanese and Chinese who share the same bloodline. Sentences such as “our own ‘super project’ connects the blood of both sides of the Strait” and “This bridge is for all the people who want to go home” are repeated across the videos.

5. Other findings from the OPORA research

Apart from shared findings, OPORA’s study identified several additional tactics observed in the Ukrainian case. 

Country-specific finding 1. Humiliation, stigmatization, and dehumanization

Much of the Russian content about Ukrainian politicians does not engage with their policies or decisions at all. Instead of criticism, it offers ridicule: AI is used to place politicians in absurd, humiliating or sexualized situations that have nothing to do with their actual work.

President Volodymyr Zelenskyy is the main target. AI-generated videos showed him in a dress dancing for Donald Trump, kissing Emmanuel Macron’s feet, or as a homeless man, a rooster or a pig. Homophobic content was especially prominent. The most recurrent plot depicted an alleged romantic or sexual relationship between Zelenskyy and Macron, told through AI videos, cartoons, allusions to films such as Twilight, and fake news segments.

The same approach extended to other politicians, including members of the opposition. Kyiv Mayor Vitali Klitschko was depicted as an “underdeveloped African,” Petro Poroshenko and Yuriy Lutsenko appeared as homeless men on the front line, and AI-generated songs described the Verkhovna Rada as a “circus” and a “theater of the absurd.” None of this content argues that a particular decision was wrong. Its effect lies elsewhere. Politicians become objects of mockery rather than participants in political debate, which may make them, and politics in general, seem unserious to viewers.

Social groups faced similar treatment, but here humiliation turns into dehumanization. Employees of the Territorial Recruitment Centers became characters in memes and AI cartoons, from a “fat TRC guy” to a scene in which a dog urinates on a TRC officer. Police officers were called “well-fed bull cops” and shown being disrespected even by children (also AI-generated). Ukrainian soldiers were depicted as weak and cowardly, for example, in a video of a frightened serviceman with a Ukrainian patch who is afraid to jump with a parachute. Stripping these groups of dignity can make hostility toward them easier to accept.

Country-specific finding 2. A large number of contradictory narratives spread simultaneously

Spreading contradictory narratives is not new for Russia. Before the AI boom, it often flooded the information space with sheer noise to confuse audiences and push them to concede that “things are not so clear-cut.” The tactic also targets many groups at the same time. 

This tactic was most clearly evident with the TikTok videos discrediting Zelenskyy. His portrayal on the platform drew on typical Russian propaganda narratives, in particular portraying the Ukrainian President as a Nazi or a Nazi collaborator. In some situations, Zelenskyy was depicted in a Nazi uniform or attributed sympathy for Nazi ideology, but in contrast, there were also other videos where he was portrayed as a Russian agent who travels to Moscow and sings the song "I am Russian," holding a USSR flag. These narratives are mutually incompatible: Zelenskyy cannot simultaneously be a Ukrainian Nazi hostile to Russia and a Russian agent loyal to Moscow. 

Other videos invoked Stepan Bandera to portray Zelenskyy as a “Banderite” and stir up hostility between Ukrainians and Poles. Such narratives do not necessarily have to form one coherent story. Different, even contradictory, portrayals can appeal to different audiences at the same time. The effect is thus broader than persuading viewers to accept one particular version of events. A large number of competing narratives can simultaneously target different groups and make the information environment itself more difficult to navigate.

Country-specific finding 3. Undermine state institutions' authority 

OPORA documented numerous attempts to portray the state’s security apparatus, including the military, the TRCs, and other law enforcement and security agencies, as repressive institutions that do not value human life. Military service was portrayed as a path to inevitable death, while mobilization was presented as a process that could only take place through violence.

However, the narratives went beyond discrediting state institutions. Some TikTok accounts also proposed ways to “solve this problem,” ranging from calls to “send the TRCs to the front” to messages normalizing violent responses to mobilization attempts. In this framing, the state’s use of coercion is presented as illegitimate and repressive, while violence against its representatives can be presented as an acceptable response.

This makes the tactic broader than simply discouraging military service or discrediting the TRCs. It challenges the legitimacy of state security institutions themselves and, in some cases, normalizes resistance to their authority through violence.

Country-specific finding 4. Deepening internal divisions

We briefly touched on identity-based manipulation when discussing the tactics shared by Russian and Chinese AI propaganda. However, the Russian case also shows a more specific approach: exploiting existing divisions within Ukrainian society and encouraging hostility between different social groups.

Ukrainian society, like any other, is not homogeneous. People have different views on political leadership, mobilization policies, and the country’s future. Russian propaganda exploits these existing disagreements and attempts to deepen hostility between groups, including military personnel, civilians, displaced people, and Ukrainians living abroad.

Among the examples documented by OPORA were videos portraying supporters of the current Ukrainian government as “pot-heads” (a derogatory nickname dating back to the Euromaidan protests, when some protesters used pots as improvised helmets) or “victory-believers,” a mocking term suggesting that they believe in a victory that may never come. These portrayals questioned the former’s ability to think critically and accused the latter of expecting others to fight on their behalf while avoiding mobilization themselves.

Similar attempts to devalue certain groups targeted Ukrainians living abroad. They were portrayed as lazy and foolish people who do not understand the realities of life in Ukraine. Other videos promoted a negative image of migrants, both in Ukraine and beyond its borders. Rather than simply reflecting existing disagreements, such content presented particular social groups as selfish, irresponsible, or undeserving of respect.

Russian propaganda seeks to turn the different, often traumatic experiences of social groups into sources of mutual distrust and hostility. Beyond amplifying existing tensions, such content portrays members of certain groups as untrustworthy and undeserving of solidarity.

Country-specific finding 5. Online content moves to real actions

OPORA identified two patterns in which AI-generated TikTok content was used not only to attract attention but also to encourage viewers to take further action.

In its study of Orthodox religious content, OPORA found that some accounts encouraged viewers to move from TikTok to Telegram or WhatsApp groups, promising opportunities to study God’s word and receive prayers and blessings. In these groups, users could communicate directly with moderators or administrators, submit requests, and receive responses or prayers (mostly in Russian and in line with the canons of the Russian Orthodox Church). This suggests that gaining likes and comments was not necessarily the final purpose of these accounts. TikTok also served as a gateway to more personal and less moderated channels of communication.

A different approach appeared in the study of pro-Ukrainian content. Some accounts used AI-generated images of Ukrainian soldiers to attract sympathy, gain likes and followers, and, in certain cases, solicit donations. Here, emotional engagement with synthetic military imagery could be turned into a request for financial support. OPORA notes, however, that such content is difficult to attribute to Russia; its harm lies in exploiting wartime grief and eroding trust in genuine fundraising.

Although these examples differ, they illustrate how AI-generated content can be used to move audiences beyond passive viewing. Rather than simply spreading narratives, such accounts encourage users to take additional steps, whether joining messaging groups or contributing money. These practices raise questions about how emotional engagement and audience trust can be exploited beyond the initial TikTok interaction.

6. Other findings from the FactLink research

Here, we also highlight findings from FactLink’s research on AI-generated TikTok videos promoting the “Xiamen-Kinmen Bridge” propaganda:

Country-specific finding 1. AI helps to visualize narratives

Propaganda not only attacks its subjects but also presents a rosy picture to its target audiences. As a tool that can quickly generate fictional images and videos, AI helps visualize narratives and strengthen propaganda’s persuasive power. As seen in Chinese propaganda videos on TikTok, AI-generated fictional images of the Xiamen-Kinmen bridge help viewers imagine the bridge and the convenience it could bring to Taiwanese people. The videos also dramatize the narrative, showing how determined the CCP is to “help the Taiwanese people.” Moreover, when fabricated information is combined with visuals, plots, and music, it can be harder for audiences to identify the problematic elements, as the many components make the content more informative and convincing. These visualized narratives pose an increasing challenge for researchers seeking to debunk them and educate the public.  

Country-specific finding 2. Longer, story-driven AI videos

As AI tools advance, they are becoming useful for creating longer, story-driven videos with complex plots, characters, music, and short-film effects. Since 2026, the FactLink team has observed a trend in Chinese propaganda toward sophisticated, drama-like AI videos. The “Xiamen-Kinmen Bridge” propaganda videos also reflect the 2026 trend of AI short dramas taking off in the Chinese-language media landscape. As AI tools become easier to use, we expect that videos lasting only a few seconds and longer drama-like videos will both become major vehicles for Chinese propaganda, combining engaging elements with subtle messages that lower audiences’ critical guard.

Country-specific finding 3. Using attractive hosts to draw attention from the audience

FactLink’s observations of AI videos themed around the “Xiamen-Kinmen Bridge” also reveal another technique used in Chinese AI propaganda: featuring attractive women discussing popular, general-interest topics that appeal to broad audiences. Using attractive women to draw viewers has been a trend across Chinese-language social media, and some of these beautiful hosts have been AI-generated. Previous investigations have also found that perpetrators use entirely AI-generated hosts or exploit stolen photos of real women to create AI avatars for propaganda and scams. In this research, FactLink found that some female hosts in the TikTok videos were AI-generated, reflecting the broader trend of using AI-created female hosts to convey problematic messages.

7. Conclusions

In this project, OPORA and FactLink teamed up to investigate how Russian and Chinese information operations use AI-generated messages to target audiences on TikTok in Ukraine and Taiwan, respectively. By comparing the tactics and narratives used by Russian and Chinese actors, we identified several features of information manipulation shared by the two authoritarian countries, alongside notable differences. These similarities and differences deepen our understanding of authoritarian information operations and help Ukraine and Taiwan anticipate potential information attacks and strengthen resilience in both societies.

In both AI-facilitated videos promoted by Russian- and Chinese-affiliated actors, researchers from Ukraine and Taiwan identify these tactics shared by Russian and Chinese propaganda:

1.       These TikTok videos combine AI-generated and authentic content, making it even harder for audiences to distinguish fabricated imagery from real scenes;

2.       Beyond using AI to create fictional plots or images, these efforts also use it to increase the volume of videos. Both OPORA and Factlink observed a growing number of shared and coordinated accounts creating and posting AI-generated videos;

3.      Russian and Chinese propaganda both use AI-generated personas to convey information. Some personas even pose as authoritative news anchors to appear credible and gain viewers’ trust;

4.      Many AI-generated videos are presented as non-political content, such as investment tips or emotional stories, to attract viewers and encourage engagement.

Researchers from Ukraine and Taiwan identify these narratives shared by Russian and Chinese propaganda:

1.       One narrative common in Russian and Chinese AI propaganda videos is: “Your government is heartless—see how good we are.” The videos depict the Ukrainian and Taiwanese governments as cruel or indifferent to their people, while presenting fabricated images of Russian leaders as always supporting their people and China as a thriving country whose government also cares about the Taiwanese; 

2.      While maligning the Ukrainian and Taiwanese governments, Russian and Chinese AI propaganda also depicts Western countries as untrustworthy allies who will either betray Ukraine or be bystanders who watch helplessly, angered by progress in China and Taiwan’s “connection”;

3.      However, both Russian and Chinese propaganda portray Ukrainians and Taiwanese people as having little choice but to accept whatever Russia and the CCP decide. Russian propaganda presents surrender as the only way to end the war, while Chinese AI propaganda insists that Taiwan must be unified and that Taiwanese people should “come home [to China].”

4.      In contrast to the denigrating content and language attacking the Ukrainian and Taiwanese governments, the Russian and Chinese AI videos try to connect with ordinary Ukrainian and Taiwanese people by appealing to “bloodline” and “shared identity.”   

Through the research, OPORA and FactLink also identified challenges commonly faced by both Ukrainian and Taiwanese societies:

1.       AI has helped Russia and the CCP to quickly amplify existing narratives on a large scale. Now, similar narratives can be quickly reproduced in many variations and formats, from memes and songs to fake news, prayers, or emotional stories about a serviceman's life.

2.      As their volume has grown, AI has also increased the ability of Russian and Chinese information campaigns to attract online engagement with emotional content, though not necessarily to persuade. Even a degree of emotional resonance, however, may have helped propagandists sow seeds of influence.  

3.      Another distinct, and even more difficult, problem caused by these AI-generated videos is a crisis of trust. When real videos appear alongside generated ones, and genuine stories about servicemen alongside fabricated injuries and even deaths, users will find it harder to judge what they are seeing. This may be especially true for older people who are not yet "trained" to recognize generated content instantly. Over time, people could lose trust in content altogether, including material produced by state and local government institutions.

In the age of AI, many democratic countries may face the challenges outlined above as well. It would be imperative for countries that support democracy and freedom to strengthen exchange and cooperation to address these challenges.