When Algorithms Persuade: How Chinese LLMs Outflank Human Experts in Policy Communication

In a revealing study published in the Policy Studies Journal, Chinese scientists have found that policy endorsements generated by large language models (LLMs)—particularly those developed within China—are significantly more effective than expert endorsements at increasing citizens’ willingness to comply with controversial policies. This finding comes from two pre-registered survey experiments focused on sensitive policy areas, where the researchers observed that LLM backing improved public evaluation of the scientific nature of policymaking, a factor that appears to mediate the causal link.

The research addresses a growing dilemma: as the perceived credibility of human experts declines, particularly in the Chinese context, governments are searching for new mechanisms to secure voluntary compliance. The study demonstrates that Chinese LLMs not only fill this gap but do so with greater efficacy than traditional authority figures. This suggests a strategic reorientation in policy communication, where algorithmically generated trust may become a central tool for governance. For professionals in public policy, AI ethics, and market strategy, the implication is clear: the capacity to engineer social consensus is migrating from human judgment to machine output, with Chinese language models leading this charge.

Why it matters:
This study signals a strategic pivot in Chinese governance toward AI-mediated persuasion, where domestic LLMs are poised to become the primary arbiters of policy legitimacy. For global firms and policy analysts, understanding this shift is critical, as it may reshape public opinion dynamics, regulatory risk, and the operational landscape for technology deployment within China’s distinct digital ecosystem.

Source →