In a noisy perceptual world, confidence judgements play a critical role in humans’ ability to make adaptive decisions in the absence of external feedback. Despite their importance, there is currently little consensus on how humans compute their confidence. To advance our understanding of these computations, we investigated the extent to which confidence judgements are informed by prior knowledge. Importantly, unlike previous research, we focused on the role of long-term priors that are formed through repeated experience with the natural world. Specifically, we used a novel psychophysical paradigm that leveraged the well-characterised probability distributions of low-level image features in natural scenes — distributions known to shape perception. In our task, participants reported the subjective upright of isolated naturalistic image regions (targets) and then reported their confidence in their orientation responses. We used computational modelling to show that participants’ perceptual and, importantly, confidence judgements aligned with those predicted by an internalised prior for natural image statistics. Our study highlights the value of naturalistic task designs that capitalise on existing, long-term priors to further understand the computational basis of decision confidence.