Generative artificial intelligence (AI) is rapidly becoming part of academic life. A March study in the UK found that 94% of undergraduate students had completed coursework assessed using generative AI technology (go.nature.com/45pced3). The impact of AI can also be seen across scientific publishing. More than half of the research papers published in 2025 contained linguistic patterns associated with large language models (LLMs), compared with approximately one in ten papers in 2023.1.
This rapid adoption of AI in education and scientific research has alarmed many educators and researchers. They fear that AI tools could weaken critical thinking, encourage students to outsource their work and overwhelm the scientific record with low-quality research. However, as a regular user of AI tools and a researcher at the University of Oxford in the UK who studies how technology influences financial-market decision-making, I believe these concerns are aimed in the wrong direction.
Gen Z’s skepticism about AI is a wake-up call — universities need to take it seriously
Effective intellectual sparring partners are difficult to find. A skilled teacher, thoughtful colleague or engaged seminar audience can challenge your assumptions, introduce evidence and respond to your ideas in real time. Yet most people rarely have access to that kind of discussion. LLMs offer, for the first time, widespread access to capable conversational partners on demand. Used well, generative AI can help distribute intellectual expertise more broadly.
Despite this potential, the dominant narrative presents AI as a threat to human reasoning. Critics argue that students are thinking less deeply and that researchers are increasingly outsourcing scientific judgment to LLMs. These risks are genuine, but the problem is not AI itself. The greater danger lies in passive AI use: treating a language model as a source of ready-made answers rather than using it to question assumptions, test ideas and improve reasoning.
Most researchers understand that scientific ideas are rarely complete when they first emerge. They develop through discussion, criticism and repeated efforts to explain them clearly. Generative AI gives scientists the opportunity to engage in these intellectual exchanges whenever and wherever they choose, while also providing a searchable record of the conversation. When used actively and critically, AI can support—not replace—human judgment, creativity and scientific thinking.
Why do universities need to fundamentally rethink exams in the AI era?
Source: www.nature.com