Social anxiety disorder (SAD) and attentional bias are theoretically connected in cognitive behavioural therapeutic models. In fact, there is an emerging field focusing on modifying attentional bias as a stand-alone treatment. However, it is unclear to what degree these attentional biases are present before commencing treatment. The purpose of this study was to measure pre-treatment attentional bias in 153 participants diagnosed with SAD using a web-based version of the dot-probe paradigm. Results showed no significant correlation for attentional bias (towards or away from negative words or faces) and the self-rated version of the Liebowitz Social Anxiety Scale (LSAS-SR). However, two positive correlations were found for the secondary measures Generalized Anxiety Disorder 7 (GAD-7) and Patient Health Questionnaire 9 (PHQ-9). These indicated that those with elevated levels of anxiety and depression had a higher bias towards negative faces in Neutral-Negative and Positive–negative valence combinations, respectively. The unreliability of the dot-probe paradigm and home-based internet delivery are discussed to explain the lack of correlations between LSAS-SR and attentional bias. Changes to the dot-probe task are suggested that could improve reliability.