Pembrolizumab is the only programmed cell death 1/programmed death-ligand 1 inhibitor for treatment of patients with non-small cell lung cancer, with a companion diagnostic assay, the 22C3 PharmDx. Although in many studies 22C3 and Ventana's SP263 appear to yield overlapping results, they show discrepancies at clinically relevant cutoffs (1% and 50%). We provide a solid comparison between 22C3 and SP263 assays in a large cohort of non-small cell lung cancer cases taking into account interobserver variability between trained pathologists who are used to either clone in their clinical practice. Serial sections of tissue microarrays, built from 198 cases of resected lung cancer, were stained for 22C3 on the Dako Link-48 platform and for SP263 on the Ventana Benchmark Ultra, following manufacturer's instructions. A protocol was also developed to run the 22C3 antibody on the Ventana platform. The pathologist used to 22C3 scored consistently higher than the pathologist used to SP263 at both 1% and 50% cutoff for all assays. For 22C3 and SP263 on respective platforms, we found statistically significant differences in terms of proportion of positive cases at both cutoffs; at 50% cutoff, around half of the cases positive with SP263 would have been defined negative with 22C3 by both pathologists. Important differences were also observed, when comparing clone 22C3 and SP263, both run on the Ventana platform. The lowest differences were seen with 22C3 run on both platforms. Assays 22C3 and SP263 show important discrepancies in identifying programmed death-ligand 1-positive cases at clinically relevant cutoffs, with possible underestimation of patients suitable for pembrolizumab therapy.