This paper states and investigates a number of problems related to the usage of XML for description of resource metadata and ontologies. The goal of the study is to elaborate a set of rules that regulate the employment of specific technology for descriptions of semantic structures. Such choice is determined by properties of the declaration language and specifics of web-resources. In the paper, the following aspects of XML-based approach are analyzed: limitations and capabilities of the language for building ontologies of different types; capabilities to extend the language for adequate knowledge representation of a subject domain; existence and employment of technologies and methods for automatic construction of metadata and ontologies; capabilities of knowledge validation on ontologies. It is stated that the XML language was not designed to be a special tool for building semantic structures. However, an XML-system (data, scheme, set of transforms) is well-placed to build ontologies or ontology-abstract software agents. In this case, the knowledge representation model can be described by a tree of concept.