BACKGROUND: Attentional bias to cannabis images is posited to drive loss of control over cannabis use and relapse in cannabis use disorder (CUD), but the literature is mixed and limited by inconsistent measurement of CUD and of confounders, including alcohol and nicotine use. This study examined attentional bias in moderate-to-severe CUD (n = 66) compared to controls (n = 42), and its relationship with cannabis/nicotine use, accounting for alcohol use. METHODS: We measured attentional bias using the visual probe task, as the difference in reaction times (RTs) for cannabis versus neutral images, in order to account for individual variability. Linear mixed effect models examined how RTs were affected by (i) group (CUD, control), image type (cannabis, neutral), group-by-image type, and group-by-image type-by-Stimulus Onset Asynchrony (SOA, 200/500 milliseconds) in the whole sample; and (ii) by image type, SOA, and moderators in the CUD group only (i.e., Cannabis Use Disorder Identification Test-Revised [CUDIT-R], subjective craving, arousal/valence ratings of the task's cannabis/neutral images, and nicotine). All models were adjusted for alcohol use. RESULTS: There were no significant group differences in attentional bias. In the CUD group, image type-by-CUDIT-R subgroups differed on RTs (β = -0.748, p =.014), whereby the high-CUDIT-R versus lower CUDIT-R subgroups had significantly faster RTs to cannabis versus neutral images (p =.034, d = -0.10), but this effect did not survive Bonferroni correction for multiple comparisons. No other results were significant. CONCLUSION: Attentional bias might not be a robust feature of CUD, though this notion requires validation in a larger sample using more direct measures of attentional bias.