Psilocybin therapy shows antidepressant potential, but its therapeutic actions are not well understood. We assessed the subacute impact of psilocybin on brain function in two clinical trials of depression. The first was an open-label trial of orally administered psilocybin (10 mg and 25 mg, 7 d apart) in patients with treatment-resistant depression. Functional magnetic resonance imaging (fMRI) was recorded at baseline and 1 d after the 25-mg dose. Beck’s depression inventory was the primary outcome measure (MR/J00460X/1). The second trial was a double-blind phase II randomized controlled trial comparing psilocybin therapy with escitalopram. Patients with major depressive disorder received either 2 × 25 mg oral psilocybin, 3 weeks apart, plus 6 weeks of daily placebo (‘psilocybin arm’) or 2 × 1 mg oral psilocybin, 3 weeks apart, plus 6 weeks of daily escitalopram (10–20 mg) (‘escitalopram arm’). fMRI was recorded at baseline and 3 weeks after the second psilocybin dose (NCT03429075). In both trials, the antidepressant response to psilocybin was rapid, sustained and correlated with decreases in fMRI brain network modularity, implying that psilocybin’s antidepressant action may depend on a global increase in brain network integration. Network cartography analyses indicated that 5-HT2A receptor-rich higher-order functional networks became more functionally interconnected and flexible after psilocybin treatment. The antidepressant response to escitalopram was milder and no changes in brain network organization were observed. Consistent efficacy-related brain changes, correlating with robust antidepressant effects across two studies, suggest an antidepressant mechanism for psilocybin therapy: global increases in brain network integration.