This study tested a 'no safe level' of alcohol consumption message on psychological outcomes relevant to drinking behaviour, compared with a gain-framed 'benefits of cutting down' message. Adults drinking at risky or harmful levels (AUDIT- ≥ 5;  = 485) were randomised to one of three messages: Control ( = 165), No Safe Level ( = 163), or Cutting Down ( = 157). Outcomes included problem recognition, defensive processing (message derogation, fear), risk recognition and self-efficacy. Compared with the control, the 'No Safe Level' message increased problem recognition (standardised mean difference = 0.22, 95% confidence interval [0.03, 0.41]), elicited stronger fear (0.92, [0.73, 1.11]), and greater derogation (0.89, [0.69, 1.09]), but reduced self-efficacy (-0.23, [-0.43, -0.02]). The 'Cutting Down' message modestly increased fear (0.30, [0.10, 0.49]) and reduced self-efficacy (-0.28, [-0.48, -0.08]) but had little effect on problem recognition or derogation. Neither message affected risk recognition. Direct comparison showed 'No Safe Level' generated more fear and derogation than 'Cutting Down', but effects on problem recognition were uncertain. No moderation by drinking level was found. Overall, the 'No Safe Level' message prompted mixed psychological responses, suggesting further attention is warranted to determine exactly what public health goals such messages may assist.