Marine top predators are often considered as sentinels of the environmental variability and bio-indicators for ecosystem structure and dynamics, given that anomalous change in their behaviour might warn us of potential marine ecosystem shifts. In this context, it is crucial to understand the multi-scale processes governing their movement and to be able to predict their response to environmental variability. Existing approach to animal movement modeling consist yet mainly in 1st order Markovian models calibrated at the local scale which can lead to overly simplistic description of trajectories. Recently, technological advances in the miniaturization of electronic devices such as GPS have enabled ecologists to document relatively large amount of animal trajectories. Parallelly, state-of-the-art tool from artificial intelligence such as Generative Adversarial Networks (GAN) have proven useful to simulate complex stochastic processes. In this study, we introduce GAN for the simulation of seabirds foraging trajectories, and we explore how GAN provide new solutions to address nowadays movement ecology challenges. More precisely, we develop a Deep Convolutional GAN (DCGAN) to simulate seabird foraging trips, and we evaluate its simulation in comparison to state-of-the-art simulation tools. Then, we investigate the ability of GANs to simulate seabird tracks from a dataset with numerous missing data using a MisGAN architecture. Finally, we explore the conditional GAN (cGAN) to predict seabirds' response to oceanographic features. We believe that GAN will open new research avenues for the exploration of ecological questions. GAN provide indeed an ultra-flexible framework where traditional methods such as state-space models struggle at predicting marine top predators' behaviour accounting for environmental heterogeneity and can suffer from computational time when maximizing likelihood of complex processes.