Background Freezing all embryos, followed by thawing and transferring them into the uterine cavity at a later stage (freeze-all), instead of fresh-embryo transfer may lead to improved pregnancy rates and fewer complications during in vitro fertilisation and pregnancies resulting from it. Objective We aimed to evaluate if a policy of freeze-all results in a higher healthy baby rate than the current policy of transferring fresh embryos. Design This was a pragmatic, multicentre, two-arm, parallel-group, non-blinded, randomised controlled trial. Setting Eighteen in vitro fertilisation clinics across the UK participated from February 2016 to April 2019. Participants Couples undergoing their first, second or third cycle of in vitro fertilisation treatment in which the female partner was aged < 42 years. Interventions If at least three good-quality embryos were present on day 3 of embryo development, couples were randomly allocated to either freeze-all (intervention) or fresh-embryo transfer (control). Outcomes The primary outcome was a healthy baby, defined as a live, singleton baby born at term, with an appropriate weight for their gestation. Secondary outcomes included ovarian hyperstimulation, live birth and clinical pregnancy rates, complications of pregnancy and childbirth, health economic outcome, and State–Trait Anxiety Inventory scores. Results A total of 1578 couples were consented and 619 couples were randomised. Most non-randomisations were because of the non-availability of at least three good-quality embryos (n = 476). Of the couples randomised, 117 (19%) did not adhere to the allocated intervention. The rate of non-adherence was higher in the freeze-all arm, with the leading reason being patient choice. The intention-to-treat analysis showed a healthy baby rate of 20.3% in the freeze-all arm and 24.4% in the fresh-embryo transfer arm (risk ratio 0.84, 95% confidence interval 0.62 to 1.15). Similar results were obtained using complier-average causal effect analysis (risk ratio 0.77, 95% confidence interval 0.44 to 1.10), per-protocol analysis (risk ratio 0.87, 95% confidence interval 0.59 to 1.26) and as-treated analysis (risk ratio 0.91, 95% confidence interval 0.64 to 1.29). The risk of ovarian hyperstimulation was 3.6% in the freeze-all arm and 8.1% in the fresh-embryo transfer arm (risk ratio 0.44, 99% confidence interval 0.15 to 1.30). There were no statistically significant differences between the freeze-all and the fresh-embryo transfer arms in the live birth rates (28.3% vs. 34.3%; risk ratio 0.83, 99% confidence interval 0.65 to 1.06) and clinical pregnancy rates (33.9% vs. 40.1%; risk ratio 0.85, 99% confidence interval 0.65 to 1.11). There was no statistically significant difference in anxiety scores for male participants (mean difference 0.1, 99% confidence interval –2.4 to 2.6) and female participants (mean difference 0.0, 99% confidence interval –2.2 to 2.2) between the arms. The economic analysis showed that freeze-all had a low probability of being cost-effective in terms of the incremental cost per healthy baby and incremental cost per live birth. Limitations We were unable to reach the original planned sample size of 1086 and the rate of non-adherence to the allocated intervention was much higher than expected. Conclusion When efficacy, safety and costs are considered, freeze-all is not better than fresh-embryo transfer. Trial registration This trial is registered as ISRCTN61225414. Funding This project was funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 26, No. 25. See the NIHR Journals Library website for further project information.