Associations of ChatGPT familiarity, student perceptions and learner digital literacy with perceived effectiveness and self-reported academic writing proficiency among university students in Mogadishu Somalia

The increasing use of generative artificial intelligence in higher education has generated growing interest in the relationship between ChatGPT and students’ academic writing skills. However, evidence from low-resource higher-education contexts is limited. This cross-sectional study examined the associations between students’ familiarity with ChatGPT (SF), students’ perceptions of ChatGPT (SP), learners’ digital literacy (LDL), perceived effectiveness of ChatGPT (EC), and self-reported academic writing proficiency (WP) among 465 undergraduate and postgraduate students enrolled in English-medium programs in Mogadishu, Somalia. Data were collected through a voluntary online questionnaire and analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). The structural model specified SF and SP as exogenous constructs, LDL as a mediating construct, and EC and WP as parallel endogenous outcomes. The direct structural paths showed that LDL was strongly associated with EC and WP, whereas SF and SP were significantly associated with LDL, EC and WP. The model explained 63%, 54%, and 69% of the variances in EC, LDL, and WP, respectively. Because academic writing proficiency was measured through self-report rather than an objective rubric-based writing assessment, the findings should be interpreted as perceived or self-reported academic writing proficiency and not as direct evidence of objectively measured proficiency. The cross-sectional design also limits causal inference. This study contributes context-specific evidence on AI-assisted writing in Somali higher education and highlights the need for artificial intelligence (AI) and digital literacy support when integrating ChatGPT into academic writing instruction.