Viral marketing is a technique that spreads advertisement information through social networks. Recently, viral marketing through online social networks has achieved huge commercial success. However, there are still very little research reported on viral marketing in decentralized mobile social networks (MSNs). Comparing with online viral marketing, viral marketing in decentralized MSNs faces many challenges, such as unreliable information diffusion and limited network knowledge. To address these problems, we propose the _gathering point-aided mobile viral marketing_ (_GP-MVM_) scheme, which contains two major component, i.e., _seed selection_ and _information diffusion_. _Seed selection_ is responsible to select a set of seed nodes from which information diffusion begins. Based on a new metric called integrated contact strength (ICS), we propose two distributed seed selection schemes, i.e., _ratio seeding_ and _threshold seeding_, while, for information diffusion, we propose the _GP-aided diffusion_ algorithm, which utilizes user GPs to promote information propagation. Continuous-time Markov chain-based analytical model shows that GP-MVM has a good scalability. Simulations indicate that GP-MVM outperforms two state-of-the-art information diffusion methods designed for MSNs, in terms of both diffusion proportion and diffusion speed.