Viruses interact with all domains of life and play fundamental roles in shaping biological systems from individual hosts to global ecosystems. Yet their identification remains difficult due to a lack of a universal marker gene and the extensive diversity of viral genomes. Despite this, the speed of viral discovery is quickly increasing, driven by the growing number of virome studies, improved sequencing technologies and the decreased cost of sequencing. In this review, we examine the evolution of virus identification approaches from classical and molecular methods to contemporary genome-resolved and computational frameworks. By aggregating genome-resolved virome studies from 2010 to early 2026 that meet defined criteria ( n =502), we synthesize the current landscape of virus identification methods, including similarity-based, sequence-based artificial intelligence (AI) and hybrid approaches. We also highlight the key limitations of the current methods, particularly biases in reference databases that contribute to persistent viral ‘dark matter’. Finally, we identify emerging opportunities for the field in structure-based and AI-driven approaches that extend detection beyond sequence similarity and outline how these integrative frameworks are poised to improve virus discovery across ecosystems.
Andrade et al. (2026) studied this question.