Background: Problem gambling is a serious public health concern at an international level where population prevalence rates average 2% or more and occurs more frequently in younger populations. The most empirically established treatments until now are combinations of cognitive and behavioural techniques labelled cognitive behaviour therapy (CBT). However, there is a paucity of high quality evidence for the comparative efficacy of core CBT interventions in treating problem gamblers. This study aims to isolate and compare cognitive and behavioural (exposure-based) techniques to determine their relative efficacy. Methods: A sample of 130 treatment-seeking problem gamblers will be allocated to either cognitive or exposure therapy in a two-group randomised, parallel design. Repeated measures will be conducted at baseline, mid and end of treatment (12 sessions intervention period), and at 3, 6 and 12 months (maintenance effects). The primary outcome measure is improvement in problem gambling severity symptoms using the Victorian Gambling Screen (VGS) harm to self-subscale. VGS measures gambling severity on an extensive continuum, thereby enhancing sensitivity to change within and between individuals over time. Discussion: This article describes the research methods, treatments and outcome measures used to evaluate gambling behaviours, problems caused by gambling and mechanisms of change. This study will be the first randomised, parallel trial to compare cognitive and exposure therapies in this population. Ethics and dissemination: The study was approved by the Southern Adelaide Health Service/Flinders University Human Research Ethics Committee. Study findings will be disseminated through peer-reviewed publications and conference presentations. Trial registration: Australian New Zealand Clinical Trials Registry: ACTRN 12610000828022.