This study examines the linguistic performance and pedagogical relevance of Qalam AI as an automatic ḥarakāt detection system in Arabic sentence learning. Employing an exploratory qualitative case study design, the research involved analysis of student text samples, expert evaluation through comparison between AI-generated outputs and manual linguistic analysis, and classroom integration simulation. The analysis focused on three grammatical cases: al-asmāʾ al-marfūʿah (nominative), al-asmāʾ al-manṣūbah (accusative), and al-asmāʾ al-majrūrah (genitive). The findings indicate that Qalam AI is capable of identifying various sentence-level linguistic features, including grammatical case assignment, orthographic inconsistencies, sentence-structure variation, and punctuation-related issues, while also exhibiting systematic limitations in contexts involving morphological ambiguity and syntactic role differentiation. Rather than functioning as an error-free automation tool, Qalam AI appears to support form-focused learning by making linguistic features visible for learner reflection and instructional mediation. These findings suggest that Qalam AI may serve as a supportive pedagogical tool within AI-assisted Arabic language instruction, complementing human linguistic judgment rather than replacing it. The study contributes to ongoing discussions in Computer-Assisted Language Learning and Arabic Natural Language Processing by highlighting the instructional value of automatic diacritization systems beyond technical accuracy.