Abstract. We consider the problem of XML schema evolution preserving the validity of existing documents related to the original schema. The aim of such schema evolution is to fit new needs without revalidating all existing valid XML documents. We propose an approach to assist users to specify schema updates that have no impact on existing document validity. An XML schema is modeled as a set of regular expressions, each constraining the content model of XML elements. Given the user needs, we work on Glushkov graphs representing regular expressions E in the schema. By this way, we get straightforwardly the right places in E that may be changed while preserving validity.