Document classification is often meant to serve as semantic indexing to help readers finding documents related to a given topic. However, the quality of indexing often deteriorates with time: some categories are misused or forgotten by indexers, others become obsolete or too general to be useful. This paper proposes measures to assess the quality of an indexing system and an algorithm that guides indexers in restructuring their indexes. Focus is put on the reader’s rather than on the annotator’s point of view (Does the classification really help accessing information? vs. Is a category adequate with the content of the document?). The whole approach is illustrated on a corpus of 20 blogs which posts are associated with categories. We show that indexers have difficulties to adapt the blogs indexing systems when the number of posts increases and we show that our approach can significantly improve the quality of these indexing systems, by simulating blog restructuring.