At least 10% of individuals who survive an acute COVID-19 infection report persistent symptoms (long COVID), which can include fatigue, neurocognitive effects and autonomic dysfunctions. However, the underlying biology is unclear; hypotheses include the persistence of SARS-CoV-2 viral antigens, autoimmunity, dysbiosis, latent virus reactivation, unrepaired tissue damage and broad immune perturbation. Reporting in Nature, Klein et al. present a cross-sectional immune-profiling study of 273 individuals that used machine learning models to identify key biological features associated with long COVID.