With an escalating number of predictive biomarkers emerging in non-small cell lung carcinoma (NSCLC), immunohistochemistry (IHC) is being used as a rapid and cost-effective tool for the screening and detection of many of these markers.In particular, robust IHC assays performed on formalin-fixed, paraffin-embedded (FFPE) tumor tissue are widely used as surrogate markers for ALK and ROS1 rearrangements and for detecting programmed death ligand 1 (PD-L1) expression in patients with advanced NSCLC; in addition, they have become essential for treatment decisions. Cytology samples represent the only source of tumor in a significant proportion of patients with inoperable NSCLC, and there is increasing demand for predictive biomarker testing on them. However, the wide variation in the types of cytology samples and their preparatory methods, the use of alcohol-based fixatives that interfere with immunochemistry results, the difficulty in procurement of cytology-specific controls, and the uncertainty regarding test validity have resulted in underutilization of cytology material for predictive immunocytochemistry (ICC), and most cytopathologists limit such testing to FFPE cell blocks (CBs). The purpose of this review is to: 1) analyze various preanalytical, analytical, and postanalytical factors influencing ICC results; 2) discuss measures for validation of ICC protocols; and 3) summarize published data on predictive ICC for ALK, ROS1, EGFR gene alterations and PD-L1 expression on lung cancer cytology. Based on our experience and from a review of the literature, we conclude that cytology specimens are in principal suitable for predictive ICC, but proper optimization and rigorous quality control for high-quality staining are essential, particularly for non-CB preparations.