Students using AI chatbots for schoolwork score lower than peers who don't, according to the OECD's latest educational report. The findings represent some of the strongest evidence yet that AI tools may harm academic performance.
The Organisation for Economic Co-operation and Development examined educational outcomes across multiple countries and found a consistent pattern: students relying on AI chatbots underperform those completing assignments independently.
The OECD's flagship report analyzed performance data and student behavior, revealing that AI use correlates with measurable drops in test scores. The research spans diverse school systems, lending weight to the conclusion that this trend extends beyond isolated cases.
Educators and researchers have raised concerns that AI chatbots may reduce the cognitive effort students invest in learning. When tools generate answers or explanations, students bypass the problem-solving process critical for developing critical thinking and subject mastery.
The findings arrive as schools worldwide grapple with AI integration. Some institutions have embraced the technology as a learning aid, while others restrict or ban student access to chatbots during academic work.
The OECD stopped short of calling for outright prohibition but emphasized the need for clear guidelines on appropriate AI use in educational settings. The report suggests that supervised, limited application may differ from unrestricted chatbot access.
Parents and administrators now face difficult decisions about classroom policy. The research doesn't address whether AI might benefit specific learning scenarios or supplementary study, focusing instead on broader patterns of academic performance.
The report adds to a growing body of evidence suggesting technology's educational impact depends heavily on implementation. Previous studies showed similar patterns with other tools, indicating that passive technology use often yields worse outcomes than active learning strategies.
Schools considering AI policies may use the OECD data to inform decisions, though researchers note that implementation details matter significantly. The next phase of study will likely examine which specific uses of AI—if any—might support rather than undermine student learning.
OpenAI has released a new AI agent that experienced unexpected behavior shortly after deployment. The incident raises fresh questions about AI safety protocols in production environments.
SpaceX AI's Grok Build tool is now available, but users planning extensive work will need to subscribe to a paid plan. The service joins a growing ecosystem of AI development tools.
Adobe is redesigning how editors access AI generators in Premiere, integrating video, audio, and music creation directly into the timeline. The overhaul aims to keep creators in their workflow rather than jumping between applications.
Rene Haas, chief executive of chip designer Arm Holdings, claims artificial intelligence will discover a cancer cure within our lifetimes. He also predicts humanoid robots could become widespread within five years.