This study investigates the alignment of AI-generated tasks with the SOLO taxonomy, a cognitive framework used in educational assessment to classify learning outcomes based on their complexity. As AI technologies increasingly contribute to educational assessments, ensuring that generated content aligns with structured frameworks like SOLO becomes essential for fostering higher-order thinking and meaningful learning. Using a combination of traditional machine learning classifiers (e.g., Multinomial Logistic Regression, Naive Bayes) and advanced models like DistilBERT, we assess the efficacy of these methods in categorizing and generating tasks across SOLO’s five levels: Pre-structural, Uni-structural, Multi-structural, Relational, and Extended Abstract. Our findings show that while traditional classifiers effectively handle lower cognitive levels (Pre-structural and Uni-structural), advanced models like DistilBERT excel across all levels, particularly at the highest cognitive level (Extended Abstract). The study addresses two research questions, focusing on the accuracy of AI-generated questions in reflecting SOLO taxonomy levels and the performance of various AI models in classifying these questions. Results indicate that Transformer-based models significantly enhance classification accuracy, especially at higher cognitive levels, offering potential advancements in AI-driven educational assessments. This research underscores the importance of aligning AI-generated assessments with established educational frameworks to enhance pedagogical validity. The insights gained provide theoretical and practical implications for designing AI-driven tools that promote critical thinking and cognitive depth in automated educational assessments.