With an unprecedented volume of data generated online by individuals, it becomes more and more relevant to analyze them for health-related purposes. However, the field of digital epidemiology lacks standardization which prevents it from moving away from a reproducibility crisis. In this article, we introduce the concept of a virtual digital cohort study (VDCS), a standardized method that can mimic an actual cohort study design exclusively thanks to social media data. We draw parallels between principles in epidemiology and their adaptations for a VDCS, and we use two diabetes-related examples to illustrate how a VDCS can help address some research questions. We also argue that developing a VDCS is a valid alternative to a costly, real-life, cohort implementation for some but not all research questions. We provide guidelines and recommendations to future researchers willing to launch a VDCS and, finally, we demonstrate why combining both a virtual and a real-life cohort study, sequentially or in parallel, can help conduct more relevant, patient-centric, research.