Résumé : Abstract Background To support the COVID‐19 pandemic response, many countries, including Belgium, implemented baseline genomic surveillance (BGS) programs aiming to early detect and characterize new SARS‐CoV‐2 variants. In parallel, Belgium maintained a sentinel network of six hospitals that samples patients with severe acute respiratory infections (SARI) and integrated SARS‐CoV‐2 detection within a broader range of respiratory pathogens. We evaluate the ability of the SARI surveillance to monitor general trends and early signals of viral genetic evolution of SARS‐CoV‐2 and compare it with the BGS as a reference model. Methods Nine‐hundred twenty‐five SARS‐CoV‐2 positive samples from patients fulfilling the Belgian SARI definition between January 2020 and December 2022 were sequenced using the ARTIC Network amplicon tiling approach on a MinION platform. Weekly variant of concern (VOC) proportions and types were compared to those that were circulating between 2021 and 2022, using 96,251 sequences of the BGS. Results SARI surveillance allowed timely detection of the Omicron (BA.1, BA.2, BA.4, and BA.5) and Delta (B.1.617.2) VOCs, with no to 2 weeks delay according to the start of their epidemic growth in the Belgian population. First detection of VOCs B.1.351 and P.1 took longer, but these remained minor in Belgium. Omicron BA.3 was never detected in SARI surveillance. Timeliness could not be evaluated for B.1.1.7, being already major at the start of the study period. Conclusions Genomic surveillance of SARS‐CoV‐2 using SARI sentinel surveillance has proven to accurately reflect VOCs detected in the population and provides a cost‐effective solution for long‐term genomic monitoring of circulating respiratory viruses.