China-linked influence operators are increasingly using artificial intelligence to target Japanese society with more sophisticated disinformation, according to research by the Sasakawa Peace Foundation and OpenAI. Generative AI is lowering barriers that once hindered Chinese information operations against Japan, particularly by helping operators overcome linguistic and cultural obstacles that made their content easy to identify.
Since Prime Minister Takaichi Sanae made comments linking a Taiwan contingency to Japan's security in late 2025, disinformation targeting Japanese political figures and domestic issues has grown in both volume and sophistication. A Nikkei investigation in February found that more than 400 China-linked social media accounts spread negative content about Takaichi during Japan's House of Representatives election. The Yomiuri Shimbun separately reported that thousands of coordinated accounts on X were involved in sharing and amplifying related content, citing research by Japan Nexus Intelligence. The narratives did not focus solely on attacking Takaichi personally but also linked Japan's defense expansion to concerns over social welfare.
Historically, Chinese influence accounts contained obvious flaws that made them detectable. Accounts were newly created, their Japanese reflected Chinese word order and linguistic habits, or AI-generated images contained visual errors. An OpenAI report released in February found that China-linked actors had used ChatGPT to assist in planning an online influence campaign targeting Takaichi. Large language models now allow operators to correct grammar, adjust tone, generate images and produce multiple narrative versions for different audiences within minutes, making China's information warfare more scalable, faster and more localized.
The Sankei Shimbun has tracked PAPERWALL, a China-linked information network operating 15 websites masquerading as Japanese news outlets. Improvements in language, imagery and the appearance of fake news sites are making it more difficult for ordinary readers to distinguish legitimate sources from fabricated ones. Language and culture can serve as tools for identifying manipulation, but even if Chinese operators use AI to reduce linguistic differences, they cannot easily reproduce the cultural background and lived experience accumulated within Japanese society.
Japan should strengthen society's ability to recognize coordinated influence operations before they gain traction rather than relying solely on fact-checking false information after it appears, experts suggest. Taiwan's experience with Chinese information operations offers a useful point of reference, with Taiwanese users identifying, reporting and blocking suspicious accounts while civil society organizations monitor information networks and patterns of coordinated behavior. Information-security cooperation between Japan and Taiwan should become more institutionalized, with deeper exchanges on identifying coordinated influence operations.
