In this study, heart rate (HR) bio‐loggers were implanted in the abdominal cavity of 12 post‐smolt Atlantic salmon Salmo salar weighing 1024 ± 31 g and acclimated to 12°C sea water. One week after the surgical procedure, a critical swim speed (Ucrit) test was performed on tagged and untagged conspecifics, whereafter tagged fish were maintained in their holding tanks for another week. The Ucrit was statistically similar between tagged and untagged fish (2.67 ± 0.04 and 2.74 ± 0.05 body lengths s−1, respectively) showing that the bio‐logger did not compromise the swimming performance. In the pre‐swim week, a diurnal cycle was apparent with HR peaking at 65 beats min−1 during the day and approaching 40 beats min−1 at night. In the Ucrit test, HR increased approximately exponentially with swimming speed until a plateau was reached at the final speed before fatigue with a maximum of 85.2 ± 0.7 beats min−1. During subsequent recovery tagged fish could be divided into a surviving group (N = 8) and a moribund group (N = 4). In surviving fish HR had fully recovered to pre‐swim levels after 24 h, including reestablishment of a diurnal HR cycle. In moribund fish HR never recovered and remained elevated at c. 80 beats min−1 for 4 days, whereafter they started dying. We did not identify a proximal cause of death in moribund fish, but possible explanations are discussed. Tail beat frequency (TBF) was also measured and showed a more consistent response to increased swimming speeds. As such, when exploring correlations between HR, TBF and metabolic rates at different swimming speeds, TBF provides better predictions. On the contrary, HR measurements in free swimming fish over extended periods of time are useful for other purposes such as assessing the accumulative burden of various stressors and recovery trajectories from exhaustive exercise.