The principle of preferentially selecting patients most likely to benefit from therapy according to their genetic profile has led to substantial clinical benefit in some tumour types, and has potential to considerably refine treatment in advanced nonsmall cell lung cancer (NSCLC). Effective, reliable use of molecular biomarkers to inform clinical practice requires the standardisation of testing methods and careful assessment of biomarkers' predictive and prognostic value. Although a number of studies have shown that patients with activating mutations in exons 18-21 of the epidermal growth factor receptor (EGFR) gene respond particularly well to gefitinib and erlotinib, a prospective, randomised study was needed to differentiate between the prognostic and predictive value of EGFR mutations. From one such study, it appeared that mutational testing should become standard at diagnosis, at least for adenocarcinoma patients with a never or low smoking history, as clinical predictors are insufficient to optimise treatment. However, outstanding questions remain: what are the treatment options for patients with tumours resistant to erlotinib/gefitinib? What conclusions about treatment can we draw from EGFR copy number or KRAS mutation status? What role should anti-EGFR antibodies play in NSCLC treatment, and in which patients? This review considers current evidence linking biomarker profile to efficacy of EGFR-targeted therapy in NSCLC, and clinical implications of recent findings.