The results of search engine usually contain a number of redundant information and how to eliminate it has become technology issues waiting to be explored. This paper proposed an improved elimination algorithm based on best similarity of the results. By analyzing the search results, extracting their key words, comparing the similarities, and clustering the results, the algorithm can eliminate the useless and reduplicate results. The experimental results show that the performance of the elimination algorithm in this paper has been much improved compared to spectral segmentation algorithm.