Abstarct: The rapid proliferation of artificial intelligence (AI) technologies has fundamentally transformed second language (L2) writing instruction, offering unprecedented opportunities for scaffolding, feedback, and content generation. However, the integration of AI into L2 writing also presents challenges, including concerns about academic integrity, over-reliance, and the potential erosion of learners' writing competence. Central to navigating this new landscape is AI literacy—the knowledge, skills, evaluation capacity, and ethical awareness necessary for effective and responsible AI use. Simultaneously, second language writing anxiety remains one of the most persistent affective barriers to EFL writing development. Despite extensive research on both AI literacy and writing anxiety, no study has examined the relationship between these constructs. This study examined the relationship between AI literacy and second language writing anxiety among Iraqi EFL university students and identified which AI literacy dimension is the strongest predictor of writing anxiety. A quantitative, cross-sectional survey design was employed with a correlational and predictive component. Data were collected from a stratified random sample of 250 Iraqi EFL university students using two validated instruments: the AI Literacy Scale (AILS; Wang et al., 2023) and the Second Language Writing Anxiety Inventory (SLWAI; Cheng, 2024 revision). Four null hypotheses were tested using Pearson correlation and multiple regression analysis. The findings revealed that Iraqi EFL students reported moderate levels of AI literacy (M = 3.40) and moderate to high levels of writing anxiety (M = 3.42). Overall AI literacy was significantly and negatively correlated with overall writing anxiety (r = -.38, p < .001). The regression model was significant, F(4, 245) = 21.45, p < .001, explaining 26% of the variance in writing anxiety (R² = .26). AI Skills was the strongest predictor (β = -.31, p < .001), followed by AI Evaluation (β = -.16, p = .013). AI Knowledge (β = -.08, p = .111) and AI Ethics (β = -.06, p = .318) were not significant predictors. Avoidance Behavior was the most strongly predicted writing anxiety dimension (R² = .23), followed by Cognitive Anxiety (R² = .17) and Somatic Anxiety (R² = .10). Students with formal AI training had significantly higher AI literacy and lower writing anxiety. These findings are consistent with Self-Determination Theory (Deci & Ryan, 1985) and Self-Efficacy Theory (Bandura, 1997). The study contributes to the theoretical understanding of AI literacy and writing anxiety in EFL contexts and provides practical guidance for AI literacy education and writing anxiety intervention in Iraq. Implications for educators, curriculum developers, and policymakers are discussed, and suggestions for future research are provided.