Abstract: An automatic abstract determination method of a document clustering is disclosed. The method includes forming a database for correspondingly packing and storing each document clustering including several abstract contents and body contents of subdocuments preloaded by a user as a document clustering process record file. Each document clustering process record file records the original abstract contents and the body contents of the subdocuments, and sets several determination values for analysis. When the user wishes to browse the contents of one document clustering, the method then determines whether or not to generate new abstract contents which is much more closer to the key points of the body contents to all of the subdocuments in the document clustering, according to the contents and values recorded in the corresponding document clustering process record file of the document clustering.