This report documents the findings and methodology of the consensus research project on Generative AI's impact on learning in K-20 U.S. education. It is intended for researchers, educators, policymakers, and other stakeholders seeking evidence-based guidance.
Generative AI (GenAI) has been described as both the solution to humanity’s problems and also ist greatest threat. In the context of education, those who look to research find mostly low-quality studies. Recent expert reports describe the risks of GenAI use but don't tie them to specific learning processes. Experts in learning and education examined where GenAI helps learning, where it harms it, and what to do about it. They identified 65 learning processes GenAI can affect and rated how much GenAI could disrupt or enhance each one. Consensus was reached on 42 of the 65 processes.

Generative AI (GenAI) has been described as both the solution to humanity's problems and also its greatest threat (Stanford Digital Economy Lab, 2026; Yudkowsky & Soares, 2025; Zuckerberg, 2026). In the context of education, policy-makers, educators, parents, and students who look to the academic literature for guidance find mostly a deluge of low-quality research (Bauer et al., 2025; Lawson et al., 2026; Weidlich et al., 2025). Recently published expert reports have focused on illuminating vulnerabilities of GenAI use in education (Burns et al., 2026; Lodge & Loble, 2026; Pearson & Jensen, 2026), but stop short of mapping them to specific learning processes or mechanisms. We sought to bring new insights to this issue by asking 30 experts in learning and education to come to consensus on what aspects of learning could be enhanced by GenAI, which aspects were vulnerable to disruption by GenAI, and what these experts recommended to promote effective learning with GenAI. What we found was sometimes surprising, often concerning, and yet hopeful for the future.

Our experts identified 65 distinct learning processes that could be affected by GenAI, and then came to consensus on the vulnerability and enhancement potential for each. This process-level granularity reveals precisely where GenAI can enhance learning and where it can disrupt it, leading to four core findings.

First, the experts' ratings revealed what we are calling the "GenAI Paradox." Casual, careless use of general-purpose GenAI (e.g., popular models from large corporations) can threaten the very learning processes necessary for committing new ideas to memory, and also the higher-order thinking needed to deeply understand and use these ideas in novel ways. However, seemingly-paradoxically, GenAI can enhance the "middle step" between initial acquisition and high-level use of new ideas: the step where learners strengthen and consolidate their initial acquisition of those ideas. In short, people should (1) initially learn without GenAI assistance, then (2) use GenAI to strengthen their initial understanding, and then (3) prioritize their own thinking when using those learned and strengthened ideas in new ways, in new contexts, and for new discoveries.

Second, the experts reached strong consensus that the disruption risk to higher-order cognitive and metacognitive processes is severe and one-sided. Specifically, excessive use of GenAI can threaten people's ability to develop critical thinking and self-regulated learning skills, or cause existing skills to decay due to not being practiced.

Third, the experts reached consensus on how GenAI can disrupt the social aspects of learning, such as human connections, social-emotional development, and collaboration skills. In an era of concern about declines in the time young people spend with each other, GenAI must not further impede the development of the social skills necessary for learning, positive relationships, life success, and civic engagement.

Fourth, among the strongest areas of consensus among our experts was that the existing empirical literature on the effects of GenAI on learning is seriously lacking in rigor and quality. Existing literature on technology's effects on learning, the small number of high-quality studies on GenAI and learning, and our experts' experiences all support the "GenAI Paradox" but there is a clear need for better research, prioritizing rigor over a rush to be "the first" to publish in this area.

To help stakeholders navigate these risks, the experts produced a prioritized recommendations action map. It begins with immediate, high-feasibility wins focused on using GenAI well and then shifts to thinking well with GenAI, prioritizing recommendations that protect students' cognitive independence, agency, and ethical norms. Finally, it highlights high-value but difficult long-term priorities that face significant systemic and institutional barriers.