Background: Social policy-makers approach the task of evidence informed policy from a perspective of ‘what is the most effective intervention to resolve a particular problem in a particular group’. Therefore, to achieve policy relevance, systematic reviews need to start from the same premise. However, the broad scope of such reviews necessitates comprehensive searches that generate high numbers of citation hits that have to be manually screened to identify relevant studies. Reviewers frequently face the problem of how to complete reviews within the available resources. This paper reports on the use of text mining to select studies for a review about engagement in culture and sport. This is the first time, as far as we are aware, that such technology has been deployed in a systematic review. Objectives: To: (i) report on a project that used text mining to support the selection of relevant studies; (ii) discuss the appropriateness and utility of this approach. Methods: Text mining technology was used to support the process of selecting relevant studies. Results: Using text mining reduced the number of potentially relevant citations that reviewers had to manually screen from 68,000 to 12,429. By using this new technology, the number of days spent identifying relevant studies was less than half the number that would have been required if we had manually screened all 68,000. Limited quality assessment suggested that the text mining was accurate at identifying studies on a topic basis, but less accurate at distinguishing research from non-research. Conclusions: *Where there are limited resources for identifying relevant studies through the traditional approach of manually screening citations, text mining may be an appropriate alternative. *Text mining does not lead to ‘bias free’ reviews, but can dramatically reduce the time taken to screen studies and produce reviews that are transparent in about what has been done.