Background Data are limited regarding the optimal dose and duration of amoxicillin treatment for community-acquired pneumonia in children. Objectives To determine the efficacy, safety and impact on antimicrobial resistance of shorter (3-day) and longer (7-day) treatment with amoxicillin at both a lower and a higher dose at hospital discharge in children with uncomplicated community-acquired pneumonia. Design A multicentre randomised double-blind 2 × 2 factorial non-inferiority trial in secondary care in the UK and Ireland. Setting Paediatric emergency departments, paediatric assessment/observation units and inpatient wards. Participants Children aged > 6 months, weighing 6–24 kg, with a clinical diagnosis of community-acquired pneumonia, in whom treatment with amoxicillin as the sole antibiotic was planned on discharge. Interventions Oral amoxicillin syrup at a dose of 35–50 mg/kg/day compared with a dose of 70–90 mg/kg/day, and 3 compared with 7 days’ duration. Children were randomised simultaneously to each of the two factorial arms in a 1 : 1 ratio. Main outcome measures The primary outcome was clinically indicated systemic antibacterial treatment prescribed for respiratory tract infection (including community-acquired pneumonia), other than trial medication, up to 28 days after randomisation. Secondary outcomes included severity and duration of parent/guardian-reported community-acquired pneumonia symptoms, drug-related adverse events (including thrush, skin rashes and diarrhoea), antimicrobial resistance and adherence to trial medication. Results A total of 824 children were recruited from 29 hospitals. Ten participants received no trial medication and were excluded. Participants [median age 2.5 (interquartile range 1.6–2.7) years; 52% male] were randomised to either 3 (n = 413) or 7 days (n = 401) of trial medication at either lower (n = 410) or higher (n = 404) doses. There were 51 (12.5%) and 49 (12.5%) primary end points in the 3- and 7-day arms, respectively (difference 0.1%, 90% confidence interval –3.8% to 3.9%) and 51 (12.6%) and 49 (12.4%) primary end points in the low- and high-dose arms, respectively (difference 0.2%, 90% confidence interval –3.7% to 4.0%), both demonstrating non-inferiority. Resolution of cough was faster in the 7-day arm than in the 3-day arm for cough (10 days vs. 12 days) (p = 0.040), with no difference in time to resolution of other symptoms. The type and frequency of adverse events and rate of colonisation by penicillin-non-susceptible pneumococci were comparable between arms. Limitations End-of-treatment swabs were not taken, and 28-day swabs were collected in only 53% of children. We focused on phenotypic penicillin resistance testing in pneumococci in the nasopharynx, which does not describe the global impact on the microflora. Although 21% of children did not attend the final 28-day visit, we obtained data from general practitioners for the primary end point on all but 3% of children. Conclusions Antibiotic retreatment, adverse events and nasopharyngeal colonisation by penicillin-non-susceptible pneumococci were similar with the higher and lower amoxicillin doses and the 3- and 7-day treatments. Time to resolution of cough and sleep disturbance was slightly longer in children taking 3 days’ amoxicillin, but time to resolution of all other symptoms was similar in both arms. Future work Antimicrobial resistance genotypic studies are ongoing, including whole-genome sequencing and shotgun metagenomics, to fully characterise the effect of amoxicillin dose and duration on antimicrobial resistance. The analysis of a randomised substudy comparing parental electronic and paper diary entry is also ongoing. Trial registration Current Controlled Trials ISRCTN76888927, EudraCT 2016-000809-36 and CTA 00316/0246/001-0006. Funding This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 25, No. 60. See the NIHR Journals Library website for further project information.