Nowadays, there is a problem in most meta-search engines that many web pages searched have nothing to do with users＇ expectations. We introduce a new user web page interactive model under the framework ofmeta search, which analyzes users＇ action to get users＇ interests and storages them, and update these information with users＇ feedback. Meanwhile this model analyzes user records stored in web, attaches labels to the web page with statistics of user interest. We calculate the similarity about user and web page with the information from model and add similarity to scores of web pages. The experimental results reveal that this method can improve the relevance of the information retrieval.