Video games could have potential therapeutic value for individuals on the autism spectrum, but little research has been done on targeting games to the diverse individual needs of adults with autism, and the problem is complicated by the inaccessibility of patient profiles. It is also important to incorporate fun as well as therapeutic value into recommendations. Fun can be estimated by comparing a user’s profile of preferred games to the proposed therapeutic games using information from online resources like VideoGameGeek and Wikipedia, even though sorting by therapeutic value is still non-trivial. This can be done by labeling therapeutic games with discrete categories according to their therapeutic value, and sorting games primarily by therapeutic category, and secondarily by estimated fun value. In this paper, we present an approach of using the patient’s profile of preferred games as a proxy for their clinical profile, and making game recommendation based on a hypothetical model and updates in response to feedback. This feedback is measured using an ad-hoc questionnaire, which is evaluated on a set of adults with autism spectrum disorder. This model both enables personalized game recommendation from a cold start and allows the learned information to be generalized to other patients.