Abstract In this paper we introduce the notion of content locality in distributed document collections. Content locality is the degree to which content-similar documents are colocated in a distributed collection. We propose two metrics for measurement of content locality, one based on topic signatures and the other based on collection statistics. We provide derivations and analysis of both metrics and use them to measure the content locality in two kinds of document collections, the well-known TREC corpus and the Networked Computer Science Technical Report Library (NCSTRL), an operational digital library. We also show that content locality can be thought of temporally as well as spatially and provide evidence of its existence in temporally ordered document collections like news feeds.