Objectives. The stages of change component of the transtheoretical model have been applied to safe sex behaviours in cross-sectional analyses, but have not yet been subject to prospective analysis. It was predicted that: (a) cross-sectional analyses would demonstrate good discrimination between the stages of change; (b) prospective analyses would allow for the identification of predictors of stage transitions; and (c) implementation intentions would explain progression from the preparation stage. Design. This study employed an experimental longitudinal design. Participants were randomly assigned to the experimental (implementation intention) or control conditions and completed questionnaires at baseline and at 2-month follow-up. Methods. Five hundred and twenty-five adolescents who were broadly representative of the UK population completed questionnaires at baseline and follow-up (N = 393) measuring: demographic variables; stage of change; theory of planned behaviour constructs; anticipated regret; and moral norm in relation to condom-carrying behaviour. The experimental condition completed a self-generated implementation intention to carry condoms at the end of the baseline questionnaire. Results. Discriminant function analyses indicated that the stages of change could be accurately discriminated from one another cross-sectionally and that, longitudinally, the measured variables were able to predict transitions between most stages. Implementation intentions caused people to progress from the preparation stage. Conclusions. Transitions between most stages were reliably predicted, thereby providing potential targets for intervention. The brief implementation intention intervention was effective and could easily be utilized and expanded to encompass a broader range of sexual health behaviours. © 2008 The British Psychological Society.