Existing solutions to data and schema integration require user interaction/input to generate a data transformation between two different schemas. These approaches are not appropriate in situations where many data transformations are needed or where data transformations have to be generated frequently. We describe an approach to an automatic XML-transformation generator that is based on a theory of information-preserving and -approximating XML operations. Our approach builds on a formal semantics for XML operations and their associated DTD transformation and on an axiomatic theory of information preservation and approximation. This combination enables the inference of a sequence of XML transformations by a search algorithm based on the operations’ DTD transformations.