Integrating GPT-4o into K-12 classrooms significantly increases students’ self-regulated learning and task engagement over time.

Objective:
The main purpose of this study was to examine the long-term effects of GPT-4o, used as an AI tutoring tool, on high school students’ self-regulated learning (SRL) and task engagement. The study sought to determine (1) how these constructs developed over the course of a semester and (2) whether students receiving AI-supported instruction showed greater growth than peers receiving traditional teaching.

Methods:
This was a longitudinal, quantitative study using a repeated-measures design. A total of 165 tenth-grade male students from a public high school were divided into two intact classes: a treatment group (n=100) that used GPT-4o for English instruction and a control group (n=65) that received teacher-centered instruction. Over a 16-week semester, both groups completed two validated surveys at three time points (Weeks 1, 8, and 16): the Self-Regulated Learning Strategy Questionnaire (SRLSQ) and the Task Engagement Questionnaire (TEQ). The treatment group interacted with GPT-4o twice weekly during 90-minute sessions to complete reading, writing, vocabulary, and grammar tasks with real-time AI feedback. Data were analyzed using latent growth curve modeling (LGCM) to examine change trajectories and repeated measures MANOVA to compare intervention effects.

Key Findings:

• Both SRL and task engagement increased significantly over the semester for all students, confirming that these are developmental constructs rather than fixed traits.

• Students in the GPT-4o treatment group showed significantly greater gains than those in the control group, demonstrating the positive impact of AI-mediated learning.

• Initial scores for SRL and engagement varied widely among students, but growth patterns were similar, suggesting AI benefited learners regardless of starting level.

• The correlation between SRL and task engagement strengthened dramatically over time, indicating that improvements in self-regulation reinforced task engagement and vice versa.

• AI-supported students experienced enhanced autonomy, competence, and relatedness—core psychological needs identified by self-determination theory—which contributed to their improved outcomes.

• Students who used GPT-4o received immediate, personalized feedback that enabled deeper reflection, more accurate self-monitoring, and greater persistence during academic tasks.

• The study found medium effect sizes for the influence of GPT-4o, indicating meaningful and educationally relevant improvements.

Implications:
The study provides strong evidence that AI-powered tools like GPT-4o can meaningfully enhance K-12 students’ self-regulation and engagement when integrated thoughtfully. By offering adaptive feedback, targeted scaffolding, and opportunities for individualized pacing, GPT-4o supported students’ psychological needs for autonomy and competence. Its integration allowed teachers to shift their attention toward higher-order instructional activities while students used the AI tool to practice skills, monitor their own learning, and revise their work. The findings suggest that AI has potential not only to improve academic performance but also to build foundational learning behaviors that are essential for long-term success.

Limitations:
The study was conducted in one all-male high school in Iran, limiting generalizability. Data relied on self-report questionnaires, which may introduce bias. The 16-week duration, while longer than many AI studies, remains too short to assess long-term persistence of effects. Additionally, only one AI tool (GPT-4o) was examined; results may differ with other AI systems.

Future Directions:
Future studies should include more diverse student populations, extended timelines, and mixed-gender environments. Researchers should incorporate observational and digital trace data to capture actual behavioral changes rather than self-perceptions alone. A broader range of AI tools should be tested to determine how different AI affordances influence SRL and engagement. Long-term follow-up research is also needed to determine whether gains continue after AI support is removed.

Title and Authors: Integrating ‘GPT-4o’ as an AI Tool Into K-12 Contexts: Assessing Its Long-Term Effects on Students' Self-Regulated Learning and Task Engagement Through Latent Growth Curve Modelling by Jie Zhang and Ali Derakhshan.

Published On: October 18, 2025 (Accepted).

Published By: European Journal of Education.