Recommendation systems have evolved significantly over the years. Companies employing these systems aim to boost their profits by suggesting customized products that users may not have initially considered but are likely to enjoy and eventually purchase, view, or try. Their applications can be observed in various sectors such as e-commerce, media, social media platforms, banking, healthcare, tourism, education, and travel. While traditional recommendation systems have a long history, their integration with Deep Learning techniques and generative models has substantially enhanced their efficiency. In this paper, we trace the evolution of recommender systems from traditional models to those employing generative models based on Deep Learning. Additionally, we delve into a qualitative comparative analysis of three generative models: VAE, GAN, and Diffusion. This study compares these algorithms on the basis of a number of criteria, including components, complexity, training process, nature of input data, data generation, requirements, and cost.