Abstract Aims The real-life benefits and risks of the non-vitamin K antagonist oral anticoagulants for stroke prevention in very elderly patients with atrial fibrillation (AF) are still debated. Methods and results Cohorts of new users of rivaroxaban 15 mg, dabigatran 110 mg, or vitamin K antagonists (VKA) for AF ≥85 years old in 2013 or 2014 were identified in the nationwide French claims database and followed-up for 1 year. Cohorts were compared after 1:1 matching using high-dimensional propensity score. Compared to VKA use and considering 1-year cumulative incidences, risk of stroke, and systemic embolism was not different with rivaroxaban use [hazard ratio 1.14, 95% confidence interval (CI): 0.93–1.40] and lower with dabigatran use (0.77, 95% CI: 0.60–0.99), risk of major bleeding was not different with rivaroxaban use (0.91, 95% CI: 0.74–1.11) and with dabigatran use (0.81, 95% CI: 0.64–1.03), risk of all-cause death was borderline to significance lower with rivaroxaban use (0.91, 95% CI: 0.83–1.00), and lower with dabigatran use (0.87, 95% CI: 0.78–0.97). The risk for a composite of all events above was not different with rivaroxaban use (0.96, 95% CI: 0.88–1.04) and lower with dabigatran use (0.87, 95% CI: 0.79–0.96) as compared with VKA use. The risk for the composite of all events was not different with rivaroxaban use as compared with dabigatran use (1.09, 95% CI: 0.97–1.23). Conclusion This study shows for the first time in more than 25 000 new real-life anticoagulant users for AF aged ≥85 years a neutral overall benefit-risk of rivaroxaban 15 mg vs. VKA and a favourable overall benefit-risk of dabigatran 110 mg vs. VKA on relevant clinical events. Study registration European Medicines Agency EUPAS14567 (www.encepp.eu) and Clinicaltrials.gov id NCT02864758.