The alignment between school textbooks and national curriculum standards is essential for effective teaching and learning, ensuring that students receive structured, relevant, and appropriate instructional material. However, in Romania, textbook evaluation has historically lacked systematic research-based validation, leading to inconsistencies in thematic coverage and readability levels. Such discrepancies result in information overload (Sweller, 1988; Spencer et al., 2019), comprehension difficulties, or instructional gaps that hinder effective classroom learning. Previous studies have identified issues such as excessive content density, inadequate thematic representation, and linguistic complexity that surpasses students’ cognitive abilities (Chitez, 2024; Oates, 2014; Valverde et al., 2002).

The primary research question guiding this study is: To what extent do Romanian school textbooks align with national curriculum standards? Our study seeks to uncover systematic discrepancies between the Romanian curriculum and correspondent textbooks based on automated linguistic analyses, thus providing a data-driven approach to textbook evaluation. Specifically, the study assesses thematic coverage using automated concept mapping, evaluates readability through computational linguistic tools, and compares textbook content with curriculum requirements via LLM-based analysis. By employing thematic mapping analysis, the study investigates whether textbooks adequately reflect the prescribed curricular priorities, ensuring that content is neither insufficient nor excessively detailed. Readability assessment provides insights into whether linguistic complexity aligns with students’ developmental stages, identifying cases where excessive syntactic density or advanced vocabulary might hinder comprehension (Mikk, 2000; McNamara et al., 2011).

The conceptual framework of this study is grounded in theories of curriculum coherence and cognitive load theory. Curriculum coherence suggests that educational materials must systematically align with national learning objectives to facilitate structured learning (Schmidt, Wang, & McKnight, 2005). Cognitive load theory (Sweller, 1988) posits that the complexity of instructional materials should match students' cognitive capacities to optimize learning. When texts contain excessive information density or advanced linguistic complexity beyond students' developmental levels, cognitive overload occurs, reducing comprehension and retention. The integration of AI-driven approaches within this framework supports an evidence-based methodology for assessing and improving textbook alignment with curriculum standards.

This study employs ROTEX-4-8, a subset of the ROTEX corpus (Romanian corpus of School Textbooks), which includes Romanian Language and Mathematics textbooks for grades 5 to 8. The analysis integrates three key AI-driven methodologies to assess curriculum-textbook alignment. Thematic mapping is conducted using Leximancer, an automated concept-mapping tool that extracts and visualizes thematic structures from textual data. By analyzing textbooks and curriculum documents, Leximancer identifies key themes, their relationships, and their distribution, revealing whether essential topics are proportionally represented, unnecessarily repeated, or omitted. Readability assessment is performed using LEMI, a Romanian-specific readability index designed to evaluate lexical and syntactic complexity in educational texts. LEMI measures linguistic features such as word length, lexical diversity, and sentence structure, providing an objective measure of whether textbook content is accessible at the intended grade level (Chitez et al., 2024). The final analytical component employs LLM-assisted content comparison to detect semantic discrepancies and alignment issues between textbooks and curricular standards. By leveraging natural language processing techniques, LLMs facilitate automated comparative analyses, identifying inconsistencies in thematic presentation and linguistic structure across different educational. Specifically, ChatGPT and Perplexity are utilized for large-scale content processing, extracting and comparing key themes, readability metrics, and structural elements across curriculum and textbook materials. These models enhance the accuracy and efficiency of curriculum alignment assessment, providing a systematic and scalable approach for evaluating instructional materials.

Our findings highlight systematic misalignments between textbooks and national curriculum standards. Thematic mapping reveals that up to 40% of the content in some subjects exceeds curricular mandates, leading to cognitive overload, while critical topics remain underrepresented. Readability analysis using LEMI indices demonstrates that Romanian Language and Mathematics textbooks for Grade 6 often surpass recommended readability levels: for instance, Flesch-Kincaid scores frequently exceed 14, Gunning Fog indexes reach over 20, and LEMI scores range between 10 and 17, far beyond the target for sixth graders. These results indicate that many instructional materials exceed students’ cognitive abilities, making comprehension difficult. Furthermore, LLM-assisted analysis identifies redundancies of 35-40% in certain textbooks (e.g., ArtKlett, Paralela 45), alongside thematic inconsistencies, where some concepts are unnecessarily complex while others are insufficiently developed. This study demonstrates the potential of AI-driven methods for evaluating textbook quality from the perspective of their alignment with curriculum standards. As educational systems increasingly seek efficient validation mechanisms, the proposed framework offers a practical solution for assessing curriculum-textbook alignment (e.g. thematic structure, readability). Its accessibility ensures that researchers, teachers, and policymakers can implement it with minimal technical expertise, making it a versatile tool for optimizing textbook evaluation. By facilitating large-scale, data-driven assessments, this framework supports evidence-based improvements in educational material design, enhancing the clarity, coherence, and accessibility of textbooks in diverse academic contexts. Given the upcoming curriculum reforms in Romania, under School Education Law 198/2023, the integration of AI-driven validation techniques is a timely and necessary step in improving the effectiveness of Romanian textbooks while also serving as a reference model for broader international applications.

Chitez, M. (2024). Linguistic overload in secondary school textbooks: A corpus-informed case study of Romanian 6th grade textbooks. In Conference Proceedings: Innovation in Language Learning 2024, Florence, Italy. Chitez, M., Dascalu, M., Udrea, A. C., Strilețchi, C., Csürös, K., Rogobete, R., & Oravițan, A. (2024). Towards Building the LEMI Readability Platform for Children’s Literature in the Romanian Language. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024) (pp. 16450-16456). McNamara, D. S., Louwerse, M. M., & Graesser, A. C. (2002). Coh-Metrix: Automated cohesion and coherence scores to predict text readability and facilitate comprehension. Technical report, Institute for Intelligent Systems, University of Memphis, Memphis, TN. Mikk, J. (2000). Textbook: Research and writing. Peter Lang. Oates, T. (2014) Why Textbooks Count: A policy paper. Cambridge: Cambridge Assessment. Schmidt, W. H., Wang, H. C., & McKnight, C. C. (2005). Curriculum coherence: An examination of US mathematics and science content standards from an international perspective. Journal of Curriculum Studies, 37(5), 525-559. Spencer, M., Gilmour, A. F., Miller, A. C., Emerson, A. M., Saha, N. M., & Cutting, L. E. (2019). Understanding the influence of text complexity and question type on reading outcomes. Reading and Writing, 32, 603-637. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285. Taga, T. (2023). What Does PISA Assess in Reading Literacy? Misconceptions and Misuses. International Journal of Education and Literacy Studies, 11(4), 57-67. Valverde, G. A., Bianchi, L. J., Wolfe, R. G., Schmidt, W. H., & Houang, R. T. (2002). According to the Book: Using TIMSS to Investigate the Translation of Policy into Practice Through the World of Textbooks. Springer Dordrecht. https://doi.org/10.1007/978-94-007-0844-0 Van Den Ham, A. K., & Heinze, A. (2018). Does the textbook matter? Longitudinal effects of textbook choice on primary school students’ achievement in mathematics. Studies in Educational Evaluation, 59, 133-140.