<![CDATA[The development of Artificial Intelligence (AI) through TikTok's auto-caption feature offers conveniences in digital communication and Arabic language learning; however, the system frequently encounters errors when processing complex Arabic linguistic structures. While previous studies have extensively discussed Arabic error analysis in academic texts and machine translation applications, research exploring real-time grammatical errors in TikTok's auto-caption feature remains limited. This study aims to analyze Arabic grammatical errors in TikTok's auto-caption feature on the @syukriarabs account through the lens of Nahwu (syntax) and Sharf (morphology). This study employs a descriptive qualitative approach with a case study design. Data were obtained from 10 videos uploaded by the @syukriarabs account using the observation and note-taking technique. The data were then analyzed using S.P. Corder’s Error Analysis theory by comparing the original audio text, the generated auto-captions, correct grammatical forms, and the specific types of errors. The results indicate that the most dominant errors occur within the phonetic and morphological domains, particularly in the misdetection of speech sounds, alterations in wazan (morphological patterns), the omission of tanwin (nunation), pronoun (dhamir) shifts, and inaccuracies in syntactic elements such as particles and i'rab (inflectional endings). Furthermore, lexical and word segmentation errors were identified, causing semantic distortion. This study concludes that TikTok's auto-caption system remains oriented toward sound-matching rather than a comprehensive understanding of Arabic linguistic context.]]>
