BACKGROUND: Procedural generation is a technique used in video game development where algorithms are used to create game assets such as landscapes and foliage. It enables game studios to create greater amounts of content than would be feasible using artists alone. For medical students, case books offer a way to apply and challenge their theoretical knowledge of diagnostics and therapy. Could procedural generation be used for generating medical cases? An unstructured search of medical literature reveals little mention of this method being explored. This project implemented and tested a limited experiment to see if the technique was applicable also in this field. METHODS: Three software tools were written: (1) A modelling tool to describe the variety in a condition s presentation, findings, lab results etc. (2) A tool that picked elements from these models in a predictable way and outputted medical cases as datasets. (3) A browser based client that presented an excerpt of a random case and asked medical students to arrive at diagnosis and treatment by requesting additional information about the patient from the case dataset. To incentivize rational choices in what information to retrieve, each request came at a cost in terms of virtual money, time and patient discomfort. After solving each case a score was presented. Students were recruited through social media to test the client tool and answer a survey about their experience. RESULTS: There were 30 respondents. Significant bias possibilities make the results unreliable. The students reported the system to be interesting (87%), useful (87%), fun (77%) and exciting (77%). 23% found the system difficult. Less than 14% found it frustrating , overwhelming , confusing or boring . Most (> 75%) perceived the medical quality to be good. The same number responded they would use such a system again, and recommended it to others, if it had more content. A similar number believed it would make them better equipped to make real diagnostic considerations, and that it would improve retention of the material studied in books. CONCLUSION: Procedural generation may be applied in medical education to generate dataset based medical cases, but significant challenges exist. Generating cases with multi- morbidity and polypharmacy appears especially difficult. Development may be costly and consume significant amounts of expert time.