The purpose of CAM research is to inform clinical CAM practice but often developments in knowledge do not translate into practice. I make three suggestions for how this situation may be improved: First, carrying out different types of study; second, enhancing dissemination; third, improving engagement. Because of the weight of prior evidence, the best interpretation of even a well-conducted positive trial of a gravely implausible CAM modality is to invoke chance. Thus no clinical trial could ever provide adequate evidence for such a modality. A way out of this problem is not to test treatment modalities for individual indications, but to test predictions of the theories of these modalities at the point at which they diverge from mainstream knowledge. Testing of this ‘basic science of CAM’ is likely to be easier, cheaper and have a larger explanatory footprint than clinical studies. Better dissemination of results can only improve the chances of research findings influencing practice, both directly by reaching practitioners, but also by reaching clients and commissioners. Suggested models include open access publishing and open data, as well as thorough use of the old and new media. Over the long term, better population health may be found through better science literacy, and the place to start is with outreach programmes in schools. In conclusion, I describe some threats and opportunities for the future and express my optimism that ineffective treatments will disappear to be replaced by effective variants which will be incorporated into evidence-based clinical practice.