There has always been an intimate and complex relationship between the diagnosis of a disease and its treatment. The approach dubbed theranostics aims to combine diagnostic techniques with therapeutic ones by deploying the same molecule in two roles, exploiting the specificity of its function to render disease treatment more effective. Does this technical development have the potential to change our conception of disease diagnosis? With the treatment approach so intimately linked to the diagnostic tool, might it be possible to treat (...) a disease without having first made an independent clinical or laboratory diagnosis? Here we discuss medical diagnosis, arguing for three categories of diagnosis, before presenting an example of a theranostic approach using radioactive prostate-specific membrane antigen ligands. This example allows us to envision a form of theranostic agent that would be able to diagnose a cancer, for example, and engage directly in its treatment, opening up the possibility of treating patients at risk of developing this cancer without any other clinical diagnostic steps. Would it be a problem if these approaches eventually became independent of any specialist clinical diagnostic supervision? If a theranostic technique is shown to work, following its own logic, do we still need an independent ‘traditional’ diagnosis prior to its use? We argue that such a diagnosis would no longer be necessary provided certain conditions are fulfilled. (shrink)