Background Evaluation of programmed cell death ligand 1 (PD‐L1) expression can be made on both resection specimens and diagnostic biopsies; however, more than 30% of patients with advanced non–small cell lung cancer (NSCLC) do not have adequate histologic material to perform PD‐L1 assays and require additional biopsies. In addition, in our practice, more than 16% of cases have cytological smears as the only available material. Our aim was to validate the PD‐L1 immunocytochemistry assay on cytological smears and compare its accuracy with the results obtained from tissue cores and whole tumor sections using the clinically relevant cutoff of 50%. Method We compared the PD‐L1 staining results of cytological smears to those from tissue cores or whole sections in 50 and 53 NSCLC cases, respectively, using the SP263 assay after scanning hematoxylin and eosin slides. Results We found an overall agreement of 90.6% between cytological smears and whole sections; specifically, we found absolute concordance between smears with PD‐L1 expressed in <10% and ≥50% of cells and whole sections with PD‐L1 expressed in <50% and ≥50% of cells, respectively. In addition, slightly lower diagnostic accuracy was found for the cytological smears in comparison with the tissue cores, but the difference was not statistically significant. We found excellent intraobserver and good interobserver agreement in the evaluation of PD‐L1 on smears. Conclusion Immunocytochemistry on cytological smears is a reliable method for determination of PD‐L1 at the 50% cutoff when positive cells are <10% or ≥50%; for cases showing PD‐L1 expression in 10% to 49% of cells, additional tissue sampling may be necessary.