Gambling self-reports may be subject to several types of bias, including social desirability bias, which may undermine their utility for capturing gambling behaviour in both research and clinical practice. Retrospective self-reports of gambling are frequently used to assess patterns of behaviour over specific periods of time, but may not be as reliable as experience sampling (ES) methods, which involve multiple assessments of gambling over the course of several days. The purpose of the current study was to examine the impact of two aspects of social desirability, impression management (IM) and self-deceptive enhancement (SDE), on the correspondence between reports of gambling assessed via ES and retrospective recall using the Gambling Timeline Followback (G-TLFB; Weinstock et al. Psychological Assessment, 16, 72–80, 2004). Participants were 81 emerging adult gamblers who completed a 30-day ES study and a retrospective assessment of their gambling. Although the overall association between social desirability and gambling reports was minimal, the correspondence between retrospective and ES reports was lower for those with higher scores on IM (for money won-lost) and SDE (for money intended to risk). Gamblers who wish to present themselves in a favourable way – either intentionally (IM) or unintentionally (SDE) – may be less reliable in their reports of gambling when asked to reflect on an extended period of time compared to when asked to provide an in-the-moment account of their gambling behaviour. These findings have important implications for understanding the circumstances under which individuals bias their retrospective self-reports of gambling and highlight the utility of more fine-grained assessments of gambling behaviour.