The medium‐term serologic response of SARS‐CoV‐2 infection recovered individuals is not well known. The aims were to quantify the incidence of seropositive failure in the medium term in a cohort of patients with different COVID‐19 severity and to analyze its associated factors. Patients who had recovered from mild and severe forms of SARS‐CoV‐2 infection in an Academic Spanish hospital (March 12–May 2, 2020), were tested for total anti‐SARS‐CoV‐2 antibodies by electrochemiluminescence immunoassay (Elecsys Anti‐SARS‐CoV‐2 test; Roche Diagnostics GmbH). The non‐seropositive status (seropositive failure) incidence (95% CI) was determined. Associations were tested by multiple logistic regression in a global cohort and severe pneumonia subpopulation. Of 435 patients with PCR‐confirmed SARS‐CoV‐2, a serological test was carried out in 325: 210 (64.6%) had severe pneumonia (hospitalized patients), 51 (15.7%) non‐severe pneumonia (managed as outpatients), and 64 (19.7%) mild cases without pneumonia. After a median (IQR) of 76 days (70–83) from symptom onset, antibody responses may not consistently develop or reach levels sufficient to be detectable by antibody tests (non‐seropositive incidence) in 6.9% (95% CI, 4.4–10.6) and 20.3% (95% CI, 12.2–31.7) of patients with and without pneumonia, respectively. Baseline independent predictors of seropositive failure were higher leukocytes and fewer days of symptoms before admission, while low glomerular filtrate and fever seem associated with serologic response. Age, comorbidity or immunosuppressive therapies (corticosteroids, tocilizumab) did not influence antibody response. In the medium‐term, SARS‐CoV‐2 seropositive failure is not infrequent in COVID‐19 recovered patients. Age, comorbidity or immunosuppressive therapies did not influence antibody response.