Spectroscopic follow-up is a pre-requisite for studies of the formation and early evolution of brown dwarfs. Here we present IRTF/SpeX near-infrared spectroscopy of 30 candidate members of the young Upper Scorpius association, selected from our previous survey work. All 24 high confidence members are confirmed as young very low mass objects with spectral types from M5 to L1, 15-20 of them are likely brown dwarfs. This high yield confirms that brown dwarfs in Upper Scorpius can be identified from photometry and proper motions alone, with negligible contamination from field objects (<4%). Out of the 6 candidates with lower confidence, 5 might still be young very low mass members of Upper Scorpius, according to our spectroscopy. We demonstrate that some very low mass class II objects exhibit radically different near infrared (0.6 - 2.5micron) spectra from class III objects, with strong excess emission increasing towards longer wavelengths and partially filled in features at wavelengths shorter than 1.25micron. These characteristics can obscure the contribution of the photosphere within such spectra. Therefore, we caution that near infrared derived spectral types for objects with discs may be unreliable. Furthermore, we show that the same characteristics can be seen to some extent in all class II and even a significant fraction of class III objects (~40%), indicating that some of them are still surrounded by traces of dust and gas. Based on our spectra, we select a sample of objects with spectral types of M5 to L1, whose near-infrared emission represents the photosphere only. We recommend the use of these objects as spectroscopic templates for young brown dwarfs in the future.