Text-mining of journal articles and other publications has long been a subject of interest. It already has applications across R&D and beyond into health care, for instance by analysing electronic health records. The technology has value but also has its limits. With new sources of text to mine becoming mainstream, such as Twitter feeds or Facebook posts that might reference a company’s brand or a drug’s efficacy or adverse events, existing technology needs to be adapted to keep pace. Not only that, but whole new compliance questions arise: does a fleeting mention on Twitter require the same response as a formal notification of an adverse event?