Cancer of unknown primary (CUP) is a metastatic malignancy for which a primary site of origin cannot be identified despite a thorough and standardized diagnostic work-up, and accounts for 1-3% of all malignancies. An unfavourable subgroup of CUP has a poor prognosis, with a median overall survival of <1 year when treated with current standard-of-care platinum-based chemotherapy. Virtually no progress in elucidating the disease biology and improving outcomes for patients with unfavourable CUP has been made over the past several decades, including a failure of initial randomized clinical trials to demonstrate the superiority of tissue-of-origin (ToO) identification by gene-expression profiling and subsequent primary-site-directed treatment over standard chemotherapy. However, large-cohort randomized trials have now shown that molecularly guided therapy improves outcomes for patients with CUP harbouring an actionable target, both in a tissue-agnostic as well as a primary tumour site-specific context. Moreover, data from non-randomized phase II trials suggest that immunotherapy using immune-checkpoint inhibitors can be beneficial even in patients with CUP that has relapsed after, or is refractory to, standard chemotherapy. In addition, a plethora of refined and novel strategies, including DNA and RNA sequencing, DNA-methylation profiling, circulating tumour DNA analysis, and artificial intelligence-based pathology, have been leveraged to facilitate ToO identification. In light of these developments, we review current ToO methodologies and compare the evidence supporting the use of a primary tumour site-guided approach versus a histology-agnostic approach to the management of CUP. We also discuss whether CUP can be viewed as a model disease for the development of histology-agnostic precision oncology treatment strategies.