BACKGROUND: Cognitive deficits across psychiatric disorders are common and difficult to treat. Semaglutide, a GLP-1 receptor agonist, shows potential cognitive benefits in diabetes and obesity, but effects in psychiatric populations remain unclear. OBJECTIVE: To determine whether initiation of semaglutide is associated with fewer cognitive signs/symptoms over 12 months among adults with psychiatric disorders. METHODS: Retrospective cohort study using de-identified multicentre US electronic health records (EHRs) in NeuroBlu Data spanning 1999-2024. We included adults with any International Classification of Diseases (ICD)-9/-10 psychiatric diagnosis with at least one semistructured, clinician-rated routine cognitive assessment in the 12 months before and at least one in the 12 months after initiating an antidiabetic drug. Semaglutide initiation was compared against four comparator strategies including sitagliptin, empagliflozin, glipizide or no antidiabetic treatment. Primary outcome was a 0-100 composite cognitive score (0, no signs/symptoms recorded; 100, signs/symptoms present across all domains) derived from semistructured, clinician-rated routine cognitive assessments across memory, attention, orientation and other cognitive functions, with a 12-month follow-up after antidiabetic drug initiation. We estimated mean ratios (MR) and 95% CIs between comparisons with the parametric g-formula, adjusting for baseline cognitive score and prespecified covariates. Additional analyses were conducted for individual cognitive domains and also grouped by major depression, psychosis and bipolar disorder. FINDINGS: Among 13 007 individuals (mean (SD) age, 62.0 (13.4) years; 6300 (48.4%) female), 1261 initiated semaglutide. Across psychiatric disorders, mean cognitive signs/symptoms scores over 12 months were lower with semaglutide (11.14) than with no treatment (14.91; MR, 0.75 (0.65-0.85); Bonferroni-corrected p=0.00002), glipizide (13.46; MR 0.83 (0.70-0.93); p=0.041) or empagliflozin (13.69; MR, 0.81 (0.69-0.94); p=0.016) and lower but not significantly different from sitagliptin (12.52; MR, 0.89 (0.77-1.02); p=0.50). Associations were stronger for memory and broader cognitive functions. Similar effect sizes were seen across disorders but were only consistently significant in major depression. CONCLUSIONS: In this EHR-based cohort of adults with psychiatric disorders, semaglutide initiation was associated with lower clinician-recorded cognitive signs/symptoms than several comparators over 12 months. CLINICAL IMPLICATIONS: Randomised clinical trials are needed to confirm causality and clinical utility of semaglutide to reduce cognitive signs/symptoms in psychiatric disorders.