Social networking and tagging have taken off at an unexpected scale and speed, opening huge opportunities to enhance the user search experience. We present Gossple, a new, user-centric, approach to improve the exploration of the Internet. Underlying \textsc{Gossple} lies the intuition that while social networks provides news from your old buddies, you can learn a lot more from people you don't know, but with whom you share many (tagging) interests. More specifically, considering a collaborative tagging system with active taggers annotating content, \textsc{Gossple} expands the search query, of any user u, with tags that are considered "close" enough with respect to users that are "close" to u. Gossple users create their own network of social acquaintances in a gossip-based manner, by dynamically computing the estimation of a distance between taggers, based on cosine similarity between tags and items. These connections are used to feed a TagMap: our central abstraction that captures the personalised relationships between tags. The TagMap is then used by Gossple} to meaningfully expand queries leveraging the personalised network. This is achieved through the TagRank algorithm, an adaptation of the celebrated pagerank algorithm, which automatically determines which tags best expand a list of tags in a given query. Gossple has no central authority: every user stores its own items and its tagging behaviour is stored only by its neighbours. The resulting networks are live, dynamic and do not require any underlying structure. We report on our evaluation of Gossple with CiteUlike traces, involving 33,834 users. In short, we show that, with little information stored at every peer, Gossple enables to retrieve items that cannot be retrieved with state of the art search systems (completeness).