The ratio of C- and M-type asymptotic giant branch (AGB) stars is commonly used to estimate the metallicity of extragalactic populations. Sources in the AGB population must therefore be accurately classified as either C- or M-type. Spectroscopic data are presented for candidate C- and M-type AGB stars, previously classified using JHK photometry, in the Local Group dwarf galaxy NGC 6822. This paper aims to evaluate the success of the JHK classification criteria used in order to determine the level of error associated with this method, and to refine the criteria for future studies. The success rate of a second independent method of source classification, the CN-TiO method, is also examined. We also review the validity of the 4kpc radial limit imposed in our previous work. Spectra of 323 sources, distributed across an area of 2deg^2^, were taken using the AAOmega multi-fibre spectrograph on the Anglo-Australian Telescope and have been classified using an automated classification system and spectral standards from the literature. Nearly half (135) of these sources were selected in common with a photometric catalogue that relied on the CN-TiO method. Within this sample we were able to classify 158 sources, including 82 C-type giants and one anomalous M-type giant, all members of NGC 6822, and 75 foreground K- and M-type dwarf sources. All but three of the giant sources are located within 3 kpc of the galactic centre. Using this spectroscopic sample, new JHK photometric criteria for the isolation and classification of C- and M-type AGB stars have been derived. The error rate in the CN-TiO method, arising from stars incorrectly classified as C-type, has been estimated to be ~7%. Based on the new JHK classification criteria, revised estimates of the global C/M ratio, 0.95+/-0.04, and iron abundance, -1.38+/-0.06dex, are presented for NGC 6822.