Non-destructively identifying the centre composition of panned chocolate goods may be useful in quality assurance settings. However, no studies to date have investigated this topic. In this study, nearinfrared spectra (1000–2500 nm) were collected from chocolate-coated peanuts and chocolate-coated sultanas ( n = 170 of each) in order to investigate the prospect of non-invasively detecting the composition of the centre. Principal component analysis confirmed that the spectra of these samples were distinct from one another. The partial least squares discriminant analysis (PLS-DA) model showed a high level of separation between chocolate-coated peanuts and sultanas in the training set (R2 = 0.95; RPD = 4.4). Discrimination between peanut and sultana samples from an independent test set was also possible, although with slightly less distinct separation between the sample types. A soft independent modelling by class analogy model was also able to differentiate between the two sample types, albeit with higher levels of misclassification compared to PLS-DA. Incorporating samples from different manufacturers may be useful for improving the broader applicability of the model.