Existing enterprises'information systems seldom take differenet requirement tendency of different personnel into consideration, the idea of Manufacturing Information Active Recommendation (MIAR) was put forward to transfer proper information to proper personnel correctly and timely. Through analyzing methods of manufacturing information acquirement in Web environment and its requirement for active recommendation service, the architecture of MIAR service was constructed. By active learning and passive learning methods, users'interest toward specific information were acquired, and content requirement structure of manufacturing information was also defined. To analyze Web pages, similarity computing model based on words distance and relevancy computing model based on key words was proposed. Manufacturing information corpus buildup by eXtensible Markup Language (XML) key words from tree structures of Document Object Module (DOM) was obtained so as to judge and collect similar Web pages according to users'interest. The MIAR was implemented by Rich Site Summary (RSS) to improve enterprises information systems'service quality and strengthen enterprises'response to dynamic information.