[REVISION HISTORY]

Impact of generative AI on Chinese student performance

Updated 2 times since CLSTR started tracking revisions of this situation.

What changed

2026-08-25 09:45 UTC → 2026-09-08 14:48 UTC · added ~~removed~~

A large-scale study tracking approximately 26,000 to 27,000 Chinese students between January 2023 and June 2025 has identified a significant discrepancy between generative AI usage and academic mastery. Research indicates that students aged 12 to 18 using AI tools, such as DeepSeek or Doubao, experienced an 18% increase in homework scores and a reduction in assignment completion time from 64 to 45 minutes. However, this efficiency did not correlate with improved testing outcomes. The study found that AI users performed 20% worse on closed-book monthly exams and saw a 24% decline in high-stakes entrance exam results compared to control groups. Researchers attribute these declines to “cognitive outsourcing,” noting that roughly 80% of users rely on AI to generate answers rather than using it as a learning aid. The negative impact on performance often manifests with a delay of up to two years, and the decline was notably more significant among top-performing students. Recent findings from a CEPR study, which tracked 26,811 students over a 30-month period, reinforce this paradox. The data confirms that while AI integration boosts homework performance, it contributes to a 24% drop in high-stakes exam results, highlighting the ongoing tension between immediate task efficiency and long-term academic retention. New data further specifies that while high-school entrance exam scores fell by 24%, college entrance exam results saw an 18% decline.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.