Automatic evaluations of clinically anxious and nonanxious children (n = 40, aged 8-16, 18 girls) were compared using a pictorial performance-based measure of automatic affective associations. Results showed a threat-related evaluation bias in clinically anxious but not in nonanxious children. In anxious participants, automatic evaluations of anxiety-relevant stimuli were more negative than those of negative stimuli. In nonanxious participants, evaluations of negative and anxiety-relevant stimuli did not differ. Furthermore, anxious youth had stronger negative evaluations of anxiety-relevant stimuli than nonanxious children. Automatic evaluations of positive, neutral, and negative stimuli did not differ between groups. Threat-related evaluations were predictive of parent-reported, but not child-reported, anxiety.