Surface-enhanced Raman spectroscopy (SERS) has been widely applied in many fields as a sensitive vibrational fingerprint technique. However, SERS faces challenges in quantitative analysis due to the heterogeneity of hot spots. An internal standard (IS) strategy has been employed for correcting the variation of hot spots. However, the method suffers from limitations due to the competitive adsorption between the IS and the target analyte. In this work, we combined the IS strategy with the 3D hybrid nanostructures to develop a bifunctional SERS substrate. The substrate had two functional units. The bottom self-assembly layer consisted of Au@IS@SiO2 nanoparticles, which provided a stable reference signal and functioned as the calibration unit. The top one consisted of appropriate-sized Au octahedrons for the detection of target analytes, which was the detection unit. Within the 3D hybrid nanostructure, the calibration unit improved the SERS performance of the detection unit, which was demonstrated by the 6-fold increase of SERS intensity when compared with the 2D substrate. Meanwhile, the reproducibility of the detection was greatly improved by correcting the hot spot changes through the calibration unit. Two biomedical molecules of cotinine and creatinine in ultrapure water and artificial urine, respectively, were sensitively determined by the 3D hybrid substrate. We expect that the developed bifunctional 3D substrate will open up new ways to advance the applications of SERS.