Mobile Ad-hoc Networks (MANETs) find numerous applications. Its utility can be extended by integrating it with Internet. Some real-life applications where such integrated networks can be used are disaster situation viz., earthquake, battle fields etc. MANETs are dynamic topology, infrastructure less and standalone networks of wireless mobile nodes. To overcome this limitation, MANET needs to connect to the Internet. In such networks, mobile nodes do not have wired connection, so they do not follow any particular route to transfer their packets. Thus, every time when topology changes, routing paths changes too. However, every node selects the shortest path to route their packet. In that case, this path will be congested due to overload. We propose a load balancing adaptive Internet gateway discovery approach which focuses on solving the problem of overload and congestion in mobile ad-hoc domain. Particle Swarm Optimization is used for selecting an optimal path among all paths available. Our work is to optimize this load balancing problem. We are using Particle Swarm Optimization (PSO) concept to overcome this problem. In our approach, we use soft computing technique viz., working procedure of PSO algorithm which searches the optimal paths using some mathematical function. Then, we combine shortest path algorithm with the gateway discovery algorithm. Thus, the algorithm first selects the shortest path stored in the memory and check whether it is free for routing packets or not, if the route is free or lightly loaded, then the source node will transfer packets otherwise if the node is overloaded, then PSO searches for the next optimal neighbour or node to route the packet. The new path is selected to route packets and stored in the place of current available path. Proposed algorithm has been implemented in MATLAB. Our approach outperforms the existing approaches.