What Research Data Actually Shows About AI in Schools

01What Research Data Actually Shows About AI in Schools

New research indicates that student adoption of artificial intelligence is occurring at a much faster pace than parents and educators realize, bringing both opportunities for personalized learning and significant risks to cognitive development. While experts warn of the dangers of technological overreliance, recent institutional initiatives are seeking to establish ethical and pedagogical guidelines to navigate this rapidly changing landscape.

02The Gap Between Use and Awareness

The integration of artificial intelligence into daily school life is already an undeniable reality, but the extent of this use frequently escapes adult oversight. A recent study conducted by researchers at the University of Southern California (USC) revealed a striking contrast: while 27% of teenagers report using AI for schoolwork multiple times a week, only 7% of parents believe their children do so with that frequency. Even more concerning is students' own perception of the technology's impact. According to Pew Research data cited in the same survey, more than half of teenagers already use AI for their assignments, and about 60% report that peers frequently use the tool to cheat. The most critical finding, however, is that a third of young people express concern that AI is impairing their ability to think for themselves. Additional analyses of students' in-school interactions with these platforms show that a full 20% of uses involve potentially troubling behavior, including cheating, bullying, or self-harm. These figures corroborate warnings issued earlier this year by the Brookings Institution, which highlighted generative AI's potential to undermine children's foundational development — affecting not only content learning but also student autonomy and agency as learners.

03The Latin American Context and Institutional Response

In Brazil and Latin America, the challenge takes on even more complex dimensions due to pre-existing structural inequalities. The region faces a severe learning crisis: six out of ten sixth-grade students do not reach minimum levels in reading and mathematics, according to the Latin American Laboratory for Assessment of the Quality of Education (LLECE). Despite this, technological adoption is advancing rapidly — more than 50% of teachers in Brazil and Chile already use AI tools, yet fewer than 10% of institutions have formal guidelines to direct this use. In response to this gap, Brazil's Ministry of Education (MEC) released in April 2026 the National Guiding Document on the Use of Artificial Intelligence for Teachers in Basic Education. The framework establishes principles grounded in human-centeredness and compliance with Brazil's General Data Protection Law (LGPD). It articulates the need to balance teaching about AI — data and algorithm literacy — with teaching with AI as a pedagogical resource, always under human supervision. Among its most forceful recommendations is the directive that all use of AI must be accompanied by teaching about AI, ensuring the technology is never treated as a black box by students. At the regional level, UNESCO launched on April 14 the Observatory on Artificial Intelligence in Education for Latin America and the Caribbean, during the Forum of the Countries of Latin America and the Caribbean on Sustainable Development, held at ECLAC headquarters in Santiago, Chile. The initiative brings together partners including Cetic.br/NIC.br (Brazil), CAF, the Ceibal Foundation, and Tecnológico de Monterrey, with the goal of generating contextualized evidence for public policy and strengthening teacher training across the region.

04Between Evidence and Implementation

The convergence of these data points and institutional responses points to a clear conclusion: artificial intelligence is neither an automatic answer to educational challenges nor an inevitable catastrophe. The real impact will depend on the pedagogical intentionality with which these tools are implemented. The USC research suggests that the most promising path for educators is to use AI for tasks that historically consume time and energy — such as repetitive grading or generating materials to differentiate instruction among students with varying proficiency levels. When technology assumes the role of a teacher's assistant, freeing educators to focus on mediation and relationship-building with students, outcomes tend to be positive. Conversely, indiscriminate use by students focused solely on delivering the final product at the expense of the learning process presents real cognitive risks. Learning requires effort, and shortcuts that eliminate the need for critical thinking can have long-term consequences for intellectual development. The central question that remains for school leaders and policymakers is not whether AI will be used in schools — it already is. The question is how to ensure its presence strengthens equity and educational excellence, rather than deepening existing gaps or compromising the intellectual autonomy of the next generation.