Epileptic seizures are currently classified based on an operational system. A complementary approach is to characterise their electrographic signature exploiting concepts from dynamical system theory. A taxonomy of sixteen classes, the dynamotypes, can be obtained by considering the bifurcations pairs allowing for the transitions between healthy and ictal state and viceversa [1]. Expanding upon this groundwork, our study aimed to validate and refine this taxonomy [2]. Our analysis was guided by insights gleaned from studying a minimal bursting model that captures different dynamotypes in a single mathematical representation [3], establishing a hierarchy among the classes of the taxonomy and the possibility that patients could exhibit more than one seizure type. We analyzed a substantial dataset comprising seizures from 120 patients with focal onset seizures, recorded via intracranial EEG. We found that data are consistent with different dynamotypes. Interestingly, we observed that individual patients may manifest different dynamotypes over time, as evidenced by analyzing longitudinal data encompassing 2000 seizures from 13 patients. Understanding patient-specific dynamotypes may hold significant implications, particularly in the context of large-scale brain models aimed at improving surgery outcomes from drug-resistant patients. For instance, the Virtual Epileptic Patient [4], currently undergoing clinical trials in France and based on a model encoding one prevalent dynamotype, may benefit from incorporating patient-specific dynamotypes, to enhance predictive accuracy.[1] Jirsa et al. 2014, Brain[2] Saggio et al. 2020, eLife[3] Saggio et al. 2017, Journal of Mathematical Neuroscience[4] Wang et al. 2023, Science Translational Medicine