The advent and popularity of the World Wide Web (WWW) has enabled access to a variety of semi-structured data and, when available, this data follows some common XML schema. On the other hand the distribution of content has made centralized solutions inappropriate, entering the era of peer-to-peer (P2P) computing, where content is stored in XML databases residing on peers. In this paper, we propose XML schema caching as a summary indexing technique for searching in P2P networks. We study XML query routing in unstructured P2P networks, comparing different search strategies and showing the advantages of our approach in terms of completeness of the search.