BACKGROUND: Increasing Watch-group antibiotic use might be contributing to antimicrobial resistance burden in sub-Saharan Africa. We evaluated the effects of a community-based, co-created intervention bundle targeting all community-level health-care providers and the communities they serve on Watch-group antibiotic use and patient management. METHODS: In a cluster-randomised, controlled trial in Nanoro, Burkina Faso, and Kimpese, DR Congo, villages or neighbourhoods with at least 500 inhabitants and at least one community-level or primary-care provider functioning as the main medicine dispenser for the population were randomly allocated (1:1) to intervention or control groups, using the RAND function in Excel. Over 9 months, three intervention rounds consisted of community health education campaigns and educational and feedback sessions with providers, introducing WHO AWaRe (Access, Watch, Reserve) Antibiotic Book guidance for infections with highest antibiotic use. We measured baseline-to-post-intervention changes in Watch-group antibiotic use through repeated patient surveys (100 per provider per village), cluster-adjusted and offset for health-care utilisation (primary outcome), and patient management scores for five common, well defined infections through simulated patient visits (secondary outcome). Providers with fewer than 20 completed surveys at baseline or post-intervention were excluded. Field workers conducting patient surveys and simulated patient visits were masked to group assignment. CABU-EICO was registered on ClinicalTrials.gov (NCT05378880). FINDINGS: 44 villages or neighbourhoods (22 each in Nanoro and Kimpese) were enrolled and randomly assigned to intervention (11 each in Nanoro and Kimpese) or control (11 each in Nanoro and Kimpese) groups. At baseline (Oct 26, 2022, to March 13, 2023), 5532 patients were surveyed (3558 in Nanoro and 1974 in Kimpese). Post-intervention (Nov 6, 2023, to April 3, 2024), 4898 patients (3180 in Nanoro and 1718 in Kimpese) were surveyed. Patients were surveyed at 32 health centres (18 intervention and 14 control), 31 private clinics (15 intervention and 16 control), 45 pharmacies (25 intervention and 20 control), and 41 informal vendors (22 intervention and 19 control). A total of 1092 simulated patient visits were completed across both periods. The weighted prevalence of Watch-group antibiotic use decreased from 26·8% (95% CI 8·8-44·8) to 17·1% (7·7-26·5) in the intervention group and increased from 13·4% (4·8-22·0) to 21·2% (8·9-33·5) in the control group; the adjusted prevalence ratio for use of Watch-group antibiotics was 0·33 (95% CI 0·14-0·78). Changes in patient management scores were minimal. INTERPRETATION: The behavioural intervention bundle was associated with a substantial reduction in Watch-group antibiotic use and no negative effect on patient management, highlighting the potential of antibiotic use improvements across health-care providers. Reduced community-level use of broad-spectrum antibiotics could help slow community-acquired pathogens' increasing resistance to clinically important antibiotics. FUNDING: The Joint Programming Initiative on Antimicrobial Resistance Research and Research Foundation-Flanders. TRANSLATIONS: For the French translation of the abstract see Supplementary Materials section.