Editor's note: Liu Yuhan, a special commentator for CGTN, is a lecturer at the Center for Pan-Himalayan Communication Studies, University of Electronic Science and Technology of China. The article reflects the author's opinions and not necessarily the views of CGTN.

As generative artificial intelligence (AI) rapidly transforms the digital landscape, the discourse on China's Xizang Autonomous Region has become increasingly vulnerable to misinformation and manipulation. The technology has dramatically lowered the barriers to fabricating and disseminating false content, enabling even ordinary users to produce highly convincing images, videos and audio materials with minimal effort.

Xizang has long been a target of international attention and even politically motivated distortions due to its unique historical, cultural and religious significance. While traditional forms of disinformation continue to circulate, AI is making such manipulation more sophisticated, more persuasive and far easier to scale.

Today, fabricated videos can create the illusion of events that never occurred, easily blurring the line between reality and fabrication; automated accounts can amplify false claims across platforms within hours; and recommendation algorithms often promote emotionally-charged content regardless of its accuracy. Together, these developments make it increasingly difficult for audiences to separate fact from fiction.

At the end of the day, what is at stake is the broader information ecosystem surrounding Xizang as misleading narratives make it hard for the world to understand the region accurately.

The rumor factory: How AI blurs the line between truth and fabrication

In the age of AI, the old adage "seeing is believing" is losing its efficacy. With only a few clicks, images and videos can be altered, and the boundary between reality and fiction deliberately obscured. Visual misinformation has become more convincing – and more deceptive – than ever before.

AI-generated misinformation concerning Xizang generally falls into two categories.

The first is cheapfake, such as repurposing unrelated footage, recycling old materials, or presenting information out of context. One example is a pseudo-documentary circulated by an overseas Facebook account which uses AI tools to alter historical footage related to Xizang and present it as authentic archival material. By presenting digital forgeries as historical evidence, such productions seek to promote selective narratives to mislead international audiences.

The second category is much more sophisticated and potentially more damaging: deepfakes. Last month, a short video circulated widely on overseas social media platforms, purporting to show Chinese authorities demolishing a Tibetan temple and expelling the monks. The scenes appeared highly realistic but subsequent verification by AFP using professional detection tools found that the video had been generated entirely by OpenAI's Sora 2, with no authentic footage whatsoever.

Multiple AI-generated images have been circulated online to give the impression of disputes surrounding religious succession and authority, serving specific political agendas and far from reflecting reality.

These targeted attempts aim to shape perceptions and manipulate public understanding. AI has not only enabled the mass production of deceptive visual content; it has also created what some scholars describe as the "liar's dividend." As authentic images become increasingly difficult to distinguish from fabricated ones, those who spread falsehoods can easily evade accountability by claiming ambiguity. Meanwhile, the public may gradually lose confidence in visual evidence altogether.

The danger is profound. In the case of Xizang, misinformation can shroud the region's history, culture and contemporary development in a fog of manufactured doubt and confusion once public trust in images and video recordings collapses.

How bot networks manufacture and magnify false narratives

While AI-generated content tools are used to produce disinformation, AI-powered bot networks are responsible for spreading it. Networks of coordinated fake accounts post in lockstep and flood the comment sections in synchronization, manufacturing the illusion of overwhelming international skepticism about China.

For example, a large number of accounts recently circulated a fabricated story claiming that France had exposed "a network of fake news sites" operated by China. The posts appeared at nearly identical times, used highly similar wording, and cross-promoted one another across the Internet to artificially boost the topic's visibility – clear indicators of coordinated bot activity.

Notably, the same tactics have been repeatedly employed. During events marking the 90th birthday of the 14th Dalai Lama, a political exile who self-claims to be a Buddhist leader, waves of automated accounts flooded social media platforms to artificially inflate public attention and engagement. When there were social media posts that the Dalai Lama had been named in documents related to Jeffrey Epstein, the US billionaire charged with sex trafficking of minors, the Dalai Lama issued a statement denying any connection to the Epstein case. Then coordinated bot accounts again moved in unison to amplify supportive narratives and shape public opinion in his favor.

Such orchestrated bot campaigns can easily distort public perception. The proliferation of repetitive, low-quality content crowds out legitimate information in platform feeds; emotionally-charged mass posting captures public attention and amplifies one-sided rumors. By creating the false impression of popular consensus, these campaigns squeeze out space for rational discussion, making it increasingly difficult for overseas audiences to get an accurate and comprehensive picture of Xizang's development.

The covert infiltrator: Biased AI LLMs construct echo chambers

Compared with overt forms of disinformation such as fabricated videos and coordinated bot campaigns, the subtle influence exerted through AI-powered large language models (LLMs) is far more difficult to detect and counter. These systems can leverage algorithms to conduct targeted and often imperceptible forms of influence, gradually shaping public perceptions and attitudes.

For example, when answering Xizang-related questions, major overseas AI models such as Claude, ChatGPT, and Grok often exhibit significant imbalances in source selection. Their responses draw heavily on Western media outlets and ideologically-driven think tanks, while rarely incorporating primary-source materials or official information from China. Even when Chinese data is cited, it is frequently sourced from secondary foreign interpretations rather than from the original authoritative sources themselves.

As a result, responses that seem to appear objective and neutral may embed implicit negative assumptions, subtly reinforcing narratives such as "China oppresses Xizang" or "Xizang remains backward."

Over time, such algorithm-driven information exposure can foster entrenched cognitive echo chambers. By continuously catering to users' existing beliefs and preferences, algorithms reinforce preconceived notions while limiting exposure to diverse perspectives. The understandings that audiences ultimately form may therefore stem not from a balanced assessment of facts, but from one-sided impressions manipulated by algorithms. Such hidden cognitive distortion is far more difficult to correct than explicit misinformation.

AI-generated false narratives about Xizang constitute a form of targeted information manipulation aimed at shaping international public opinion. Their purpose is to disrupt the global information environment, weaken China's discourse power on Xizang-related issues, and discredit China's policies on the governance and development of Xizang.

Addressing this new generation of AI-driven information operations requires a coordinated response. Digital platforms must strengthen efforts to curb false and misleading content; official institutions should communicate a comprehensive account of Xizang's development; and the public must improve its ability to critically assess AI-generated information and identify misinformation.

Only by ensuring that technology serves the public good can the disorder of the information space be effectively contained. Ultimately, facts remain the most powerful tool for exposing and defeating AI-generated falsehoods.

(If you want to contribute and have specific expertise, please contact us at opinions@cgtn.com. Follow @thouse_opinions on X to discover the latest commentaries in the CGTN Opinion Section.)