The structure of a network has an effect on how well the information in the network can be ranked, finds a study in Nature Communications this week. The work uses a popular algorithm used to rank web content by search engines and could have implications for how the results of ranked information are used in science and marketing.

Pagerank is used to rank web content by search engines such as Google. The algorithm counts each link in a network as a vote and ranks the information. Gourab Ghoshal and Albert-Laszlo Barabasi show that the structure of a network affects its performance. They also conclude that Pagerank may be inherently more accurate for some networks than for others. They find that exponential networks, such as a food web, are prone to perturbations. For scale-free networks, such as the web, the growth of information and content available improves the ranking by making the top ranks more obvious and stable.