Null mutations in the UGT1A1 gene result in Crigler–Najjar syndrome type I (CNSI), characterized by severe hyperbilirubinemia and constant risk of developing neurological damage. Phototherapy treatment lowers plasma bilirubin levels, but its efficacy is limited and liver transplantation is required. To find alternative therapies, we applied AAV liver-specific gene therapy to a lethal mouse model of CNSI. We demonstrated that a single neonatal hUGT1A1 gene transfer was successful and the therapeutic effect lasted up to 17 months postinjection. The therapeutic effect was mediated by the presence of transcriptionally active double-stranded episomes. We also compared the efficacy of two different gene therapy approaches: liver versus skeletal muscle transgene expression. We observed that 5–8% of normal liver expression and activity levels were sufficient to significantly reduce bilirubin levels and maintain lifelong low plasma bilirubin concentration (3.1±1.5 mg/dl). In contrast, skeletal muscle was not able to efficiently lower bilirubin (6.4±2.0 mg/dl), despite 20–30% of hUgt1a1 expression levels, compared with normal liver. We propose that this remarkable difference in gene therapy efficacy could be related to the absence of the Mrp2 and Mrp3 transporters of conjugated bilirubin in muscle. Taken together, our data support the concept that liver is the best organ for efficient and long-term CNSI gene therapy, and suggest that the use of extra-hepatic tissues should be coupled to the presence of bilirubin transporters.