Background There has been increasing interest in the potential for pre‐emptive interventions in the prodrome of autism, but little investigation as to their effect. Methods A two‐site, two‐arm assessor‐blinded randomised controlled trial (RCT) of a 12‐session parent‐mediated social communication intervention delivered between 9 and 14 months of age (Intervention in the British Autism Study of Infant Siblings‐Video Interaction for Promoting Positive Parenting), against no intervention. Fifty‐four infants (28 intervention, 26 nonintervention) at familial risk of autism but not otherwise selected for developmental atypicality were assessed at 9‐month baseline, 15‐month treatment endpoint, and 27‐ and 39‐month follow‐up. Primary outcome: severity of autism prodromal symptoms, blind‐rated on Autism Observation Schedule for Infants or Autism Diagnostic Observation Schedule 2nd Edition across the four assessment points. Secondary outcomes: blind‐rated parent–child interaction and child language; nonblind parent‐rated communication and socialisation. Prespecified intention‐to‐treat analysis combined estimates from repeated measures within correlated regressions to estimate the overall effect of the infancy intervention over time. Results Effect estimates in favour of intervention on autism prodromal symptoms, maximal at 27 months, had confidence intervals (CIs) at each separate time point including the null, but showed a significant overall effect over the course of the intervention and follow‐up period (effect size [ES] = 0.32; 95% CI 0.04, 0.60; p = .026). Effects on proximal intervention targets of parent nondirectiveness/synchrony (ES = 0.33; CI 0.04, 0.63; p = .013) and child attentiveness/communication initiation (ES = 0.36; 95% CI 0.04, 0.68; p = .015) showed similar results. There was no effect on categorical diagnostic outcome or formal language measures. Conclusions Follow‐up to 3 years of the first RCT of a very early social communication intervention for infants at familial risk of developing autism has shown a treatment effect, extending 24 months after intervention end, to reduce the overall severity of autism prodromal symptoms and enhance parent–child dyadic social communication over this period. We highlight the value of extended follow‐up and repeat assessment for early intervention trials.