Adaptive Generative Tutoring in Resource-Limited Secondary Education: Effects on Participation, Achievement, and Digital Equity in Nigeria

This study examines the short-term educational contribution of teacher-guided generative tutoring through a secondary analysis of a randomized trial in nine public secondary schools in Benin City, Nigeria. Among 1328 randomized students, 657 were offered twelve 90 min sessions combining Microsoft Copilot, curriculum-aligned English activities, and pair work, while 671 received regular instruction. Intention-to-treat models used ordinary least squares, school fixed effects, baseline performance adjustment, and HC3 standard errors, with conditional estimates on observed endline outcomes and required covariates. The overall assessment effect was 0.310 SD (95% CI 0.176–0.444; n = 654), with positive estimates for English (0.238 SD), AI knowledge (0.309 SD), and school examination performance (0.206 SD). The digital-skills estimate was positive but imprecise (0.139 SD; p = 0.066). Attendance was positively associated with achievement, and effects were larger at higher baseline performance. The female interaction was positive, with caution warranted by the smaller male subsample. The digital-access interaction was negative but imprecise (−0.182 SD; 95% CI −0.370–0.005; 457 complete cases). These findings support short-term learning gains from the complete tutoring package relative to regular instruction. They provide a basis for further evaluation of structured programs integrating generative AI with curriculum, collaborative practice, and teacher guidance in resource-limited urban schools.