The severity of coronavirus disease 2019 (COVID-19) may be influenced by pre-existing immune responses against endemic coronaviruses, but conflicting data have been reported. We studied 148 patients who were hospitalised because of a confirmed diagnosis of COVID-19, classified mild in 58, moderate in 44, and severe in 46. The controls were 27 healthy subjects. At admission, blood samples were collected for the measurement of biomarkers of disease severity and levels of the IgG against the receptor-binding domain (RBD) of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) and pre-existing coronaviruses OC43, HKU1, NL63 and 229E. Higher levels of IgG antibodies against the RBD of pre-existing coronavirus (with the highest significance for anti-HKU1 IgG, p = 0.01) were found in patients with mild disease, compared with those with moderate or severe disease. Multivariable logistic regression confirmed the association of high levels of antibodies to pre-existing coronavirus with mild disease and showed their associations with low levels of the complement activation marker SC5b-9 (p range = 0.007–0.05). High levels of anti-NL63 antibodies were associated with low levels of the coagulation activation marker D-dimer (p = 0.04), while high levels of IgG against 229E were associated with low levels of the endothelial activation marker von Willebrand factor (p = 0.05). Anti-SARS-CoV-2-neutralising activity of plasma positively correlated with anti-SARS-CoV-2 IgG (r = 0.53, p = 0.04) and with anti-HKU1 IgG (r = 0.51, p = 0.05). In hospitalised patients with COVID-19, high levels of antibodies to pre-existing coronaviruses are associated with mild disease, suggesting that their measurement could be useful in predicting the severity of the disease.