Tumours with high somatic mutation rates escape immune surveillance by upregulating receptors and ligands such as programmed death receptor-1 and its ligand (PD-1/PD-L1). Checkpoint inhibitors (ICI) provide encouraging therapeutic results in non-small cell lung cancers (NSCLC) and may soon be used in 2nd or 1st line therapy. Currently PD-L1 immunohistochemistry (IHC) expression assessed on tumour cells is used as a predictive biomarker, since better patient outcomes are often, but not always associated with increased tumour cell PD-L1 IHC expression. However pre-analytical variables, different anti-PD-L1 clones used on different staining platforms, different specimens types, as well as intra- and interobserver variability influence the results. We will only understand PD-L1 expression on tumour cells if we accept that PD-L1 is an inducible pathophysiological factor with variable levels of PD-L1 expression depending on the immunological status. Should we test PD-L1 during initial diagnostic work up before, or at the point when immune checkpoint therapy is considered? Taking all arguments into account the value of PD-L1 as a predictive biomarker is questionable. Other predictive biomarkers such as high mutation burden, mRNA expression, neo-antigens and the diversity of tumour antigen-specific T cells should be evaluated in the future. Here we review results presented in 30 journal articles and three reviews covering this topic in the last 3 years.