Cancer immunotherapy utilizing immune checkpoint inhibitors, has become mainstay for a growing number of cancer types. However, in general, the objective response rate and survival is only improved in a minority of cancer patients. Many of the biomarkers to select patients that either benefit or do not, have next to being predictive also prognostic value, are continuous instead of categorical, and are therefore difficult to use in daily practice. Currently, PD-L1 by immunohistochemical staining is the only approved biomarker to select metastatic non-small cell lung cancer patients for single agent pembrolizumab treatment. This may be extended with PD-L1 IHC staining in other cancer types, with different assays and different cut-offs. Other biomarkers, such as tumor mutational burden (TMB) by either whole exome sequencing or by targeted panel sequencing, gene expression profiling (GEP) either by NanoString or another RNA based sequencing test are under development. Assessment of immune infiltrates by multiplex immunohistochemistry or immunofluorescence also requires further validation. In the end, some of these tests may need to be combined in order to have the highest positive and negative predictive value and qualify as a reliable test for patient selection. The complexity and current status of biomarker research in immuno-oncology is outlined in this review.