The glucose oxidase–peroxidase (GOD–POD) method used to measure serum unbound bilirubin (UB) suffers from direct bilirubin (DB) interference. Using a bilirubin-inducible fluorescent protein from eel muscle (UnaG), a novel GOD–POD–UnaG method for measuring UB was developed. Newborn sera with an indirect bilirubin/albumin (iDB/A) molar ratio of <0.5 were classified into four groups of DB/total serum bilirubin (TB) ratios (<5%, 5–10%, 10–20%, and ≥20%), and the correlation between the UB levels and iDB/A ratio was examined. Linear regression analysis was performed to compare UB values from both methods with the iDB/A ratio from 38 sera samples with DB/TB ratio <5% and 11 samples with DB/TB ratio ≥5%. The correlation coefficient (r) between UB values and the iDB/A ratio for the GOD–POD method was 0.8096 (DB/TB ratio <5%, n = 239), 0.7265 (5–10%, n = 29), 0.7165 (10–20%, n = 17), and 0.4816 (≥20%, n = 16). UB values using the GOD–POD–UnaG method highly correlated with the iDB/A ratio in both <5% and ≥5% DB/TB ratio sera (r = 0.887 and 0.806, respectively), whereas a low correlation (r = 0.428) occurred for ≥5% DB/TB ratio sera using the GOD–POD method. Our GOD–POD–UnaG method can measure UB levels regardless of the presence of DB.