Antithrombin deficiency, the most severe thrombophilia, might be underestimated, since it is only investigated in cases with consistent functional deficiency or family history. We have analyzed 444 consecutive, unrelated cases, from 1998 to 2021, with functional results supporting antithrombin deficiency in at least one sample. Plasma antithrombin was evaluated by functional and biochemical methods in at least two samples. SERPINC1 gene was analyzed by sequencing and MPLA. Hypoglycosylation was studied by electrophoresis and high‐performance liquid chromatography (HPLC). In 260 of 305 cases (85.2%) with constitutive deficiency (activity < 80% in all samples), a SERPINC1 (N = 250), or N‐glycosylation defect (N = 10) was observed, while 45 remained undetermined. The other 139 cases had normal antithrombin activity (≥ 80%) in at least one sample, what we called transient deficiency. Sixty‐one of these cases (43.9%) had molecular defects: 48 had SERPINC1 variants, with two recurrent mutations (p.Ala416Ser[Cambridge II], N = 15; p.Val30Glu[Dublin], N = 12), and 13 hypoglycosylation. Thrombotic complications occurred in transient deficiency, but were less frequent, latter‐onset, and had a higher proportion of arterial events than in constitutive deficiency. Two mechanisms explained transient deficiency: The limitation of functional methods to detect some variants and the influence of external factors on the pathogenic consequences of these mutations. Our study reveals a molecular defect in a significant proportion of cases with transient antithrombin deficiency, and changes the paradigm of thrombophilia, as the pathogenic effect of some mutations might depend on external factors and be present only at certain timepoints. Antithrombin deficiency is underestimated, and molecular screening might be appropriate in cases with fluctuating laboratory findings.