It is considerably harder to generalize identity across different pictures of unfamiliar faces, compared to familiar faces. This finding hints strongly at qualitatively distinct processing of unfamiliar face stimuli – for which we have less expertise. Yet the extent to which face selective vs. generic visual processes drive outcomes during this task has yet to be determined. To explore the relative contributions of each, we contrasted performance on a version of the popular ‘Telling Faces Together’ unfamiliar face matching task, implemented in both upright and inverted orientations. Furthermore, we included different age groups [132 British children aged 6 to 11-years (69.7% Caucasian), plus 37 British Caucasian adults] to investigate how participants’ experience with faces as a category influences their selective utilisation of specialised processes for unfamiliar faces. Results revealed that unfamiliar face matching is highly orientation-selective. Accuracy was higher for upright compared to inverted faces from 6 years of age, which is consistent with selective utilisation of specialised processes for upright vs. inverted unfamiliar faces during this task. The effect of stimulus orientation did not interact significantly with age, and there was no graded increase in the magnitude of inversion effects observed across childhood. Still, a numerically larger inversion effect in adults compared to children provides a degree of support for developmental changes in these specialised face abilities with increasing age/experience. Differences in the pattern of errors across age groups are also consistent with a qualitative shift in unfamiliar face processing that occurs some time after eleven years of age.