Autobiographical memory is typically held to comprise episodic and semantic elements, with the vast majority of studies to date focusing on profiles of episodic details in health and disease. In this context, ‘non-episodic’ elements are often considered to reflect semantic processing, or are discounted from analyses entirely. Mounting evidence suggests that rather than reflecting one unitary entity, semantic autobiographical information may contain discrete subcomponents, which vary in their relative degree of semantic or episodic content. The current study aimed to (i) review the existing literature to formally characterise the variability in analysis of ‘non-episodic’ content (i.e., external details) on the Autobiographical Interview and (ii) use these findings to create a theoretically-grounded framework for coding external details. Our review exposed discrepancies in the reporting and interpretation of external details across studies, reinforcing the need for a new, consistent approach. We validated our new external details scoring protocol (the ‘NExt’ taxonomy) in patients with Alzheimer’s disease (n = 18) and semantic dementia (n = 13), and 20 healthy older Control participants, and compared profiles of the NExt subcategories across groups and time periods. Our results revealed increased sensitivity of the NExt taxonomy in discriminating between autobiographical memory profiles of patient groups, when compared to traditionally used internal and external detail metrics. Further, remote and recent autobiographical memories displayed distinct compositions of the NExt detail types. This study is the first to provide a fine-grained and comprehensive taxonomy to parse external details into intuitive subcategories, and to validate this protocol in neurodegenerative disorders.