Sleep disturbances are present in ~65% of individuals with generalised anxiety disorder (GAD). Although both Kundalini yoga (KY) and cognitive behavioural therapy (CBT) are effective treatment options for GAD, little is known about how these treatments compare in improving sleep for GAD and what drives these changes. Accordingly, we examined the effects of CBT, KY, and stress education (SEdu; an attention control condition) on subjective sleep quality (as measured by the Pittsburgh Sleep Quality Index [PSQI] and Insomnia Severity Index [ISI]) in a randomised controlled trial of 226 adults with GAD (mean age 33.37 years; 70% female; 79% White). We hypothesised that both CBT and KY would outperform SEdu in improving sleep disturbances. Three potential mediators of sleep improvement (worry, mindfulness, perceived stress) were also examined. In line with hypotheses, PSQI and ISI scores significantly improved from pre- to post-treatment for all three treatment groups (all p < 0.001, all d > 0.97). However, contrary to predictions, sleep changes were not significantly greater for CBT or KY compared to SEdu. In mediation analyses, within-person deviations in worry, mindfulness, and stress each significantly mediated the effect of time on sleep outcomes. Degree of change in sleep attributable to worry (CBT > KY > SEdu) and perceived stress (CBT, KY > SEdu) was moderated by treatment group. Personalised medicine as well as combined treatment approaches should be studied to help reduce sleep difficulties for patients with GAD who do not respond.