There is a common mischaracterisation that autistic individuals have reduced or absent empathy. Measurement issues may have influenced existing findings on the relationships between autism and empathy, and the structure of the empathy construct in autism remains unclear. The present study sought to address these gaps by examining the structure and psychometric properties of the Perth Empathy Scale (PES) in autistic individuals (N = 239) compared to non-autistic individuals (N = 690). Our moderated non-linear factor analysis revealed that the multidimensional empathy construct manifested similarly in autistic and non-autistic individuals, with the PES displaying good validity and reliability. Moreover, the results revealed that autistic individuals reported reduced cognitive empathy and reduced affective empathy for positive and negative emotions. However, there was greater heterogeneity of empathic tendencies in the autistic sample, indicating that these mean differences may not be generalisable for all autistic individuals. The present study highlights that the PES is suitable for assessing empathy across autistic and non-autistic individuals. This work with the PES also provides greater nuance to our understanding of empathy and autism, and based on these findings, we propose the empathy heterogeneity hypothesis of autism as a new way of describing empathy in autism.