Focusing on the case of medically unexplained symptoms (MUS), this article explores diagnostic classification in the absence of biomedical evidence or other strong medical warrants for diagnosis. The data are from three focus group interviews with Norwegian general practitioners (GPs) conducted in 2015, that centred on the issue of what diagnoses to use (or not) for MUS. The qualitative analysis reconstructs the logic underlying GPs’ diagnostic accounts, which centred on the meaning of diagnostic categories and on anticipating how ‘generalised others’ would respond to those meanings (called ‘diagnosing by anticipation’). The analysis suggests that GPs confer diagnoses by balancing unwarranted medical accuracy and anticipated harmful diagnostic consequences; the goal of diagnosis was finding categories in the International Classification of Primary Care that would yield acceptable results, without making a liar of the GP in the process. Drawing on the distinction between diagnosis as colligation and classification, the findings and their relevance for medical sociology are discussed. Counter to frequent descriptions as ‘illness that cannot be diagnosed’, the analysis shows how GPs can diagnose MUS in the bureaucratic sense of diagnosis as classification – a sense that has been missing from sociological view.