Predictive prospective studies of schizophrenia date back to the late 1950s. At the turn of the Millennium, an Australian research group initiated programs of early detection of schizophrenia and early therapeutic intervention. The theoretical foundations of early schizophrenia detection usually remain unaddressed. In this paper, we focus on the issue of prediction of future schizophrenia in the general population on the basis of a symptomatic picture. Several notions used in this research program such as disease, symptom, and clinical staging derive from a medical model, which in our view is not entirely adequate for grasping the nature of schizophrenia. Schizophrenia is a spectrum of disorders with a shared core Gestalt comprising dis-order of selfhood and intersubjectivity. This core Gestalt has manifold manifestations, often predominantly in the existential or experiential domain. It is not feasible to apply medical concepts to this symptomatically poorly demarcated spectrum for which we do not know robust biological validators. Moreover, there is increasing evidence that the current DSM and ICD criteria of schizophrenia distort the original concept of schizophrenia and are formulated on a very high severity level. This often results in incorrect diagnoses of young help-seeking patients. In sum, it seems more appropriate to detect psychosis among already help-seeking patients than to implement detection programs in the general population. We discuss a reorientation of psychiatry towards more refined psychopathological knowledge and assessment that are needed in order to optimize the treatment of young help-seeking patients.