In terms of dual process models, behaviour can be conceived of as the outcome of an interplay between reflective, top-down and impulsive, bottom-up processes. Behaviour change interventions may benefit from targeting both types of processes in a coherent way. One approach to this, in the context of reducing hazardous drinking, is to combine imagery involving real-life situations involving alcohol with the simple actions involved in Approach Bias Modification (ApBM), a form of Cognitive Bias Modification. We developed and tested a version of this Imagery-enhanced Approach Bias Modification (IApBM) in an experimental design, with two independent factors: imagery versus control and ApBM versus control training components (N = 139). An effect of integrating the training factors was hypothesized on the alcohol-approach bias of an alcohol Approach-Avoidance Task. Further exploratory analyses were performed for the bias on alcohol-related Single Attribute Implicit Association Tests and on alcohol-related questionnaires. Finally, the psychometric properties of an imagery interference effect during training were explored. Results showed no benefit of the training and in fact suggested a negative interaction in which combining the training components appeared to block reductions in craving effected by each in isolation. The reliability of the imagery-related interference effect was high and the effect was correlated with alcohol-related scales. In conclusion, it appears that interference between training components decreases their individual effects when combining imagery and ApBM in the current way. The imagery-related interference effects that could be measured during training conditions may be useful as an implicit measure of automatic processes underlying hazardous drinking. [Abstract copyright: © 2022 British Psychological Society.]