Introduction: AI, or artificial intelligence, is appearing in K-16 classrooms for lesson planning, automated quizzes and grading, etc. Although gains in instructional efficiency and assessment reliability have been reported, there are still few comparative studies using a single AI across different policy, curriculum, and infrastructure contexts from various countries. Purpose: Assess the influence of Notegrade.ai on (i) lesson-planning efficiency and curriculum alignment, (ii) quiz generation quality, and (iii) grading speed, reliability, and perceived fairness by users in several different national educational systems. Study Design: This was a multi-site assessment using a mixed-methods concurrent triangulation design. We quantified pre/post adoption time spent on task (lesson planning, quiz writing, grading) and AI–human agreement in grading open- and closed-response questions (using Cohen’s κ/ICC) to analyze agreement between AI and human graders. We examined curriculum against rubric-based ratings that were linked to outcomes and levels of Bloom’s taxonomy. Fairness was assessed through item-level differential item functioning (DIF) analysis and through subgroup error analysis. Surveys based on TAM/UTAUT constructs, semi-structured interviews, classroom observations, and artifact audits were used to gather teacher and student perceptions. Multilevel regression with site-level covariates of digital readiness, assessment policy, connectivity, and teacher workload norms was used to model cross-national heterogeneity. Results: Teachers at all sites experienced a significant time savings while using Notegrade.ai to prepare lesson plans and quizzes which they used to focus more on feedback and differentiation. AI-based grading was found to have high inter-rater reliability for objective items, as well as moderate-to-high reliability for rubric scored answers and significantly reduced turnaround time. Rubric checks revealed that the generated quizzes perfectly aligned with the desired outcomes and included items at all levels of cognitive complexity. No systematic bias in subgroups was identified in the fairness analyses, although some variations were found in conditions with low connectivity and minimal training of teachers. National policy supports and specificity in local curriculum moderated adoptive as well as perceived useful . Discussion: If accompanied by score-matching rubrics, rater-calibration methods, and some rudimentary training of teachers, Notegrade.ai may have a positive impact on classroom productivity without compromising assessment quality in diverse educational contexts. AI use must be coupled with physical and training support; developers must allow visible controls for aligning content to standards, calibrating rubrics, and detecting bias. Recommendations for Practice and Research: The next steps should involve longitudinal studies that connect efficiencies made possible through the use of AI technology with actual student learning, studies that replicate this model in low resource environments, as well as cost-effectiveness research that account for total cost of ownership.
Akinwumi Fakokunde (Fri,) studied this question.