Randomized clinical trials (RCTs) are an important source of evidence for clinical practice, but finding and applying RCT reports to care is time consuming. Publishing RCTs directly into machine-understandable “trial banks” may allow computers to deliver RCT evidence more selectively and effectively to clinicians. Methods: Authors of eligible RCTs published in JAMA or the Annals of Internal Medicine between January 2002 and July 2003 were invited to co-publish their trial in RCT Bank, an electronic knowledge base containing details of trial design, execution, and summary results. Trial bank staff used Bank-a-Trial, a web-based trial-bank entry tool, to enter information from the manuscript into RCT Bank, obtaining additional information as necessary from the authors. Results: The author participation rate rose from 38% to 76% after the first co-published trial was available as an example. Seven diverse RCTs are now co-published, with 14 in progress. Conclusions: We have demonstrated proof of concept for co-publishing RCTs with leading journals into a structured knowledge base. Phase II of trial bank publishing will introduce direct author submission to RCT Bank.