Background: It is not yet clear whether nutritional status is associated with post-stroke cognitive impairment (PSCI). We examined the geriatric nutritional risk index (GNRI) on the domain-specific cognitive outcomes 3 months after a stroke. Methods: A total of 344 patients with acute ischemic stroke were included for the analysis. The GNRI was calculated as 1.489 × serum albumin (g/L) + 41.7 × admission weight (kg)/ideal body weight (kg) and was dichotomized according to the prespecified cut-off points for no risk and any risks. The primary outcome was PSCI, defined as having adjusted z-scores of less than −2 standard deviations in at least one cognitive domain: executive/activation, memory, visuospatial and language. Multiple logistic regression and linear regression analyses were performed to investigate the association between the GNRI and cognitive outcomes. Results: Seventy (20.3%) patients developed PSCI 3 months after a stroke. The mean GNRI was 106.1 ± 8.6, and 59 (17.2%) patients had low (<98) GNRI scores. A low GNRI was independently associated with the PSCI after adjusting for age, sex, education, initial stroke severity, stroke mechanism and left hemispheric lesion (odds ratio, 2.04; 95% confidence interval, 1.00–4.14). The GNRI scores were also significantly associated with the z-scores from the mini-mental status examination and the frontal domain (β = 0.04, p-value = 0.03; β = 0.03, p-value = 0.03, respectively). Conclusions: A low GNRI was independently associated with the development of PSCI at 3 months after an ischemic stroke. The GNRI scores were specifically associated with the z-scores of the global cognition and frontal domain cognitive outcomes.