Despite neurobiological overlap, alcohol use disorder (AUD) and body mass index (BMI) show minimal genetic correlation (rg), possibly due to mixed directions of shared variants. We applied MiXeR to investigate shared genetic architecture between AUD and BMI, conjunctional false discovery rate (conjFDR) to detect shared loci and their directional effect, Local Analysis of (co)Variant Association (LAVA) for local rg, Functional Mapping and Annotation (FUMA) to identify lead single nucleotide polymorphisms (SNPs), Genotype-Tissue Expression (GTEx) to examine tissue enrichment, and BrainXcan to assess associations with brain phenotypes. MiXeR indicated 82.2% polygenic overlap, despite a rg of -.03. ConjFDR identified 132 shared lead SNPs, with 53 novel, showing both concordant and discordant effects. GTEx analyses identified overexpression in multiple brain regions. Amygdala and caudate nucleus volumes were associated with AUD and BMI. Opposing variant effects explain the minimal rg between AUD and BMI, with implicated brain regions involved in executive function and reward, clarifying their polygenic overlap and neurobiological mechanisms.