Glioblastoma, IDH-wildtype (GBM), is the most common primary brain cancer and is associated with a poor prognosis. Anti-PD-L1 immune checkpoint inhibitors have shown promising results in other solid tumors, prompting studies evaluating their potential in GBM. Accurately determining programmed death-ligand 1 (PD-L1) expression, using different antibody clones, is essential for assessing the potential efficacy of immunotherapy in GBM patients. However, the variability in immunohistochemical expression of different PD-L1 clones in GBM has not been tested. This study aims to assess and compare PD-L1 expression using two antibody clones SP263 and 22C3 in a retrospective GBM cohort. PD-L1 testing yielded spatial variability within the same tumor sample and significant discordance between the tested PD-L1 assays with the SP263 clone consistently detecting higher PD-L1 positivity. Overall, 46.5% of GBM cases were PD-L1 positive (≥ 5%) with SP263, compared to 32.6% with 22C3. The inter-clone kappa agreement reached 0.71. Positive cases (≥ 5%) were 53.8% and 38.5% in the unmethylated MGMT GBM subgroup with SP263 and 22C3 respectively, while 35.3% or 23.5% in the methylated MGMT GBMs. Our findings underscore that commonly used PD-L1 assays yield discordant results, highlighting the need to establish a gold standard for PD-L1 testing as a predictive biomarker.