Abstract Background Blood eosinophil counts (BEC) were recently included in the 2019 Global Initiative for Obstructive Lung Disease (GOLD) guideline as an easily accessible theragnostic biomarker for Chronic Obstructive Pulmonary Disease (COPD). However, the stability of BEC remains insufficiently studied. Methods We conducted a retrospective study in six primary care practices in Belgium on data from Electronic Health Records of stable COPD patients, to characterise the stability of blood eosinophils over time. We report the percentage of patients with BEC persistently below or above the 2019 GOLD guideline thresholds (100 and 300 cells/μL). For each patient the mean, standard deviation (SD) and relative standard deviation (RSD) of the BEC were calculated to determine the intra-patient variability. Results Ninety-eight patients were included, yielding 1082 eosinophil measurements (median 8 measurements/patient), with BEC ranging between 0 and 1504 cells/μL. Four (4.1%) patients had BEC persistently below 100 cells/μL, 34 (34.7%) had measurements persistently above this threshold. Approximately half of the patients (51.0%) had BEC persistently below 300 cells/μL and 3 (3.1%) patients had counts persistently above this threshold. 28.6% of patients crossed both threshold values throughout the registration period. The mean BEC per patient ranged between 15 and 846 cells/μL with an intra-patient SD between 5 and 658 cells/μL. The mean intra-patient RSD was 0.46. There was a significant strong positive correlation (Pearson analyses) between the mean BEC and SD (r = 0.765; n = 98). Simple linear regression was used to further describe the influence of the mean eosinophil count on the SD (B = 0.500; 95%CI 0.415–0.586; n = 98; p < 0.001). Conclusion BEC can be variable in individual COPD patients. Therefore, the use of a single measurement to guide therapeutic decisions remains debatable. Further prospective research remains necessary to validate the reproducibility of this biomarker.