Computer-assisted learning is an approach which has been used widely to support people with autism spectrum disorders. Commercial and research-driven technologies continue to be produced at a rapid rate, particularly mobile device applications. However, the field lacks a consistent, evidence-based methodology for design, implementation and evaluation of these technologies. This review collates published evidence for the value of CAL in autism education and therapy, with a specific view to identifying best practice in design, implementation and evaluation. The future of research in CAL for autism is considered including the need for relevant theoretical underpinnings, and appropriate responses from researchers and practitioners to novel technologies.