Mothers of very premature newborns often have low milk supply. Systematic review has shown increased milk quantity with relaxation interventions. We hypothesised that a self-directed audio relaxation and lactation-specific visualisation would increase milk quantity after a very premature birth. Unmasked, randomised, controlled trial, recruiting 132 participants in four United Kingdom neonatal units. Eligible women had given birth to one or two infants between 23+0 and 31+6 weeks of gestation. The intervention was a 12-min voice recording including breathing exercises, muscle relaxation and lactation-specific visualisation. Primary outcome was the highest 24-h breastmilk weight expressed on any of day 4, day 14 or day 21 after birth. Mean birth gestation was 27.8 weeks (SD 2.4), with 26% of participants giving birth under 26 weeks (34/132). Adjusted mean difference in primary outcome was 73.9 g (95% CI -61.7 to 209.5, p = 0.28). Spielberger State-Trait Anxiety Index adjusted mean difference was -1.9 (-8.2 to 4.3, p = 0.54). The majority of relaxation group participants felt the intervention was relaxing (32/42, 76%). There was no beneficial effect of this relaxation intervention on milk quantity. Mothers of very premature infants may value relaxation interventions but they are unlikely to have a large effect on milk quantity. This randomised trial did not show a beneficial effect of a self-directed audio relaxation and visualisation on mothers' own milk quantity expressed after very preterm birth. Mothers of very and extremely preterm infants may value relaxation interventions, but they are unlikely to have a large effect on milk quantity. Prior systematic review of mixed populations has shown an increase in mothers' own milk quantity with relaxation interventions. Combining this study with existing meta-analysis could result in a new hypothesis that the lower the gestation at birth, the smaller the impact of relaxation on milk quantity.