With the exponential growth of data presentation in recent years, it has become an important and urgent need to quickly extract and integrate key contents from a large number of documents. Currently, there are approaches to optimize the summary extraction of multiple documents from a single source. However, in practical applications, documents often come from multiple sources, so the problem faced is transformed into a summary extraction problem from multiple source documents. To address the issues, this paper proposes a multi-source and multi-document summary extraction optimization algorithm based on association and clustering (M2SEO-AC). In data preprocessing, the clustering algorithm is utilized to divide documents into groups to obtain representative documents from each source so that all important contents can be effectively kept in the evolution process. The designed fitness function considers not only the coverage, reducing redundancy, content relevance, and association centrality but also the diversity factor to evaluate the quality of a summary. The goal of the diversity factor is to ensure that contents from sources are balanced in the final summary, preventing a single source from dominating the content of the summary, thereby providing a more balanced and unbiased summary. In the experimental results, the DUC2002 dataset is used to verify the proposed approach. The results show that with the addition of the diversity factor in the fitness function, compared with the existing methods, the proposed M2SEO-AC approach has improved the ROUGE score performance and effectively extracted the representativeness of the contents of the summary.