The relationship between thiamine blood level (TBL) and cognition remains uncertain, including among alcohol-dependent persons (ADP). To evaluate this relationship during protocol-driven inpatient alcohol detoxification treatment including thiamine supplementation (AD + Th). Prospective 3-week study with 100 consecutively admitted detoxification-seeking ADP (47.7 ± 11 years old, 21% females) without superseding comorbidities requiring treatment. TBL and Montreal Cognitive Assessment (MoCA) were measured at admission (t 1 , pre-AD + Th) and discharge (t 3 , post-AD + Th). Frontal Assessment Battery (FAB) was performed at t 1. AD + Th included abstinence, pharmacological alcohol withdrawal syndrome treatment, and oral thiamine supplementation (200 mg/day for 14 days). Regression and mediation analyses assessed TBL-cognition relationships. We found no cases of Wernicke Encephalopathy (WE) and only one case of thiamine deficiency. Both MoCA and TBL significantly improved across AD + Th (with medium-to-large effect sizes). At t 1 , TBL significantly predicted MoCA and FAB sum scores (medium effect sizes; extreme and very strong evidence, respectively). The clear TBL-MoCA association disappeared at t 3. In multivariate regression and mediation analyses exploring key influential factors of cognition (identified by LASSO regression), the TBL-MoCA interactions did not relevantly change at t 1 and t 3. Age, serum transaminases, vitamin D levels, drinking-years, and depression score weakly modified the relationship. TBL was a robust predictor of pre-detoxification cognitive impairment, and both TBL and cognition improved significantly during AD + Th (including abstinence) in our ADP population, supporting routine thiamine supplementation for ADP, even those at low WE-risk. The TBL-cognition relationship was minimally confounded by age, alcohol-toxicity proxies, mood, and vitamin D levels. • The relationship between thiamine blood level (TBL) and cognition is still unclear. • This is a prospective 3-week study with 100 detoxifying adult alcohol-dependent persons. • TBL and cognition significantly improved during the detoxification treatment including thiamine supplementation. • TBL significantly predicted pre-detoxification cognition, esp. visuospatial, executive, memory, and abstraction abilities. • Significant TBL-cognition association was missing at post-detoxification (presumably owning to a ceiling effect upon MOCA). • TBL-cognition association was not meaningfully confounded by alcohol-toxicity proxies, homocysteine, mood, serum vitamin D. [ABSTRACT FROM AUTHOR]