Academics are increasingly embracing the use of artificial intelligence (AI) in educational settings, with a recent study revealing a shift in attitudes toward AI use in university courses across the United States. Conducted by Igor Chirikov, an education researcher at the University of California, Berkeley, the study analyzed over 31,000 course syllabi from a large public university in Texas and found that restrictive policies implemented following the launch of ChatGPT in late 2022 have softened across most disciplines, with the exception of arts and humanities.
Chirikov’s research, published in Berkeley’s open-access repository, utilized a large language model (LLM) to analyze syllabi from 2021 to 2025. This task, which would have taken human coders approximately 3,000 hours, showed a notable transition towards more permissive AI policies by the autumn of 2025. In the spring of 2023, discussions around academic integrity were mentioned in 63 percent of course materials. By fall 2025, that figure had dropped to 49 percent, indicating a broader acceptance of AI’s role in education.
In contrast to earlier concerns, policies are now increasingly focused on requiring students to attribute their AI use. At the beginning of 2023, only 1 percent of syllabi included such stipulations, but by the end of 2025, this number had risen to 29 percent, demonstrating a significant shift in perspective. While references to AI as a learning tool remain limited, they increased to 11 percent by fall 2025, up from nearly zero.
Chirikov noted that course materials are evolving to include explicit guidelines on AI usage, with a notable distinction based on the task at hand. For tasks involving drafting or revising, 79 percent of policies still prohibit AI assistance. Conversely, in coding or technical work, only 20 percent banned AI use, and for editing or proofreading, the figure was 17 percent. This nuanced shift suggests that academics are utilizing their “professional judgments” during a critical period of AI adoption.
During a recent interview with Times Higher Education, Chirikov remarked that many scholars appear to be “warming to” and “experimenting with” AI rather than outright banning it. He emphasized that instructors are actively redesigning courses to incorporate assignments that require AI use, which goes against the notion that faculty are simply relinquishing control over AI’s role in education. Indeed, some courses are now treating AI as a tool that students should utilize throughout the curriculum, including during assessments.
This shift challenges the prevailing narrative that academics favor blanket bans on AI usage. Chirikov explained that outright prohibitions are often impractical and difficult to enforce. Instead, instructors are increasingly specifying acceptable AI use for particular tasks, aligning their policies with actual learning processes. He argues that the conversation surrounding AI’s role in education is shifting from concerns over academic integrity to discussions on how AI can enhance the learning experience.
Chirikov believes that while academic integrity remains important, the broader focus should be on the skills students are developing and the learning practices they are engaging in. More instructors are beginning to frame AI in terms of learning, considering when it can support educational practices and when it might replace essential skills students need to acquire.
The findings from this study reflect a significant evolution in how higher education is approaching the integration of AI technologies. As the academic landscape continues to adapt, the implications for teaching and learning processes could reshape educational paradigms, paving the way for a more nuanced understanding of AI’s potential in fostering student engagement and enhancing educational outcomes.
See also
AI Tools Boost Scientific Output by 36% but Risk Monoculture, New Studies Reveal
AI Lab Notebooks Transform Research Workflows, Boosting Hypothesis Generation and Analysis
Cisco Launches G300 AI Infrastructure with 33% Network Efficiency Boost at Cisco Live EMEA
UC San Diego Reveals AI Training Method Surpassing 85% Accuracy in Multimodal Reasoning
Machine Learning Breakthrough Reveals Core Sequence in Aptamer Discovery, Enhancing Binding Affinity
