The browsing behavior of massive web users forms a flow network transporting user’ collective attention between websites. By analyzing the circulation of the collective attention we discover the scaling relationship between the impact of sites and their traffic. We construct three clickstreams networks, whose nodes were websites and edges were formed by the users’ switching between sites. The impact of site i, C i , is measured by the clickstreams controlled by this site in the circulation of clickstreams. We find that C i scales sublinearly with A i , the traffic of site i. Specifically, there existed a relationship C i ~ A i γ (γ < 1), which implies the decentralized structure of the clickstream circulation.