Expression of programmed cell death-ligand 1 (PD-L1) is being used as predictive biomarker for immunotherapy in head and neck squamous cell carcinoma (HNSCC). Several antibodies are available for PD-L1 testing and multiple staining and scoring methods are used. This study aimed to compare the performance of two PD-L1 standardized assays (SP263 and 22C3 pharmDx) and one laboratory-developed test (LDT) (22C3) in HNSCC using the tumor proportion score (TPS) and the combined positive score (CPS). Pretreatment biopsies from 147 HNSCC patients were collected in a tissue-microarray (TMA). Serial sections of the TMA were immunohistochemically stained for PD-L1 expression using 22C3 pharmDx on the Dako Link 48 platform, SP263 on the Ventana Benchmark Ultra platform, and 22C3 as an LDT on the Ventana Benchmark Ultra. Stained slides were assessed for TPS and CPS. Cutoffs of ≥1% and ≥50% for TPS and ≥1 and ≥20 for CPS were used. Concordance between the different staining assays was moderate to poor for TPS (intraclass correlation coefficient (ICC) 0.46) as well as for CPS (ICC 0.34). When stratifying patients by clinically relevant cutoffs, considerable differences between the assays were observed: concordance was poor for both TPS and CPS. Generally, SP263 stained a higher percentage of cells than the other assays, especially when using the CPS. Moderate concordance was shown between three different PD-L1 immunohistochemical assays and considerable differences in PD-L1 positivity were observed when using clinically relevant cutoffs. This should be taken into account when using PD-L1 expression to guide clinical practice.