Background: Long COVID (LC) is a multisystem clinical syndrome with functional disability and compromised overall health. Information on LC clinical severity types is emerging in cross-sectional studies. This study explored the pattern and consistency of long COVID (LC) clinical severity types over time in a prospective sample. Methods: Participants with LC completed the condition-specific outcome measure C19-YRSm (Yorkshire Rehabilitation Scale modified version) at two assessment time points. A cluster analysis for clinical severity types was undertaken at both time points using the k-means partition method. Results: The study included cross-sectional data for 759 patients with a mean age of 46.8 years (SD = 12.7), 69.4% females, and a duration of symptoms of 360 days (IQR 217 to 703 days). The cluster analysis at first assessment revealed three distinct clinical severity type clusters: mild (n = 96), moderate (n = 422), and severe (n = 241). Longitudinal data on 356 patients revealed that the pattern of three clinical severity types remained consistent over time between the two assessments, with 51% of patients switching clinical severity types between the assessments. Conclusions: This study is the first of its kind to demonstrate that the pattern of three clinical severity types is consistent over time, with patients also switching between severity types, indicating the fluctuating nature of LC.