This paper addresses the significant challenges that the uncertainty of renewable energy (RE) outputs, such as wind and solar power, bring to distribution network planning and operation by proposing a multi-objective bi-level distribution network planning model based on an improved generative adversarial network and carbon footprint analysis. Firstly, a Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) is employed to simulate numerous wind and solar output scenarios, which are then reduced using the K-medoids clustering algorithm. Secondly, carbon footprint coefficients for each generation unit are determined through a life cycle assessment method. Next, a bi-level distribution network planning model considering carbon footprints is established: the upper level minimizes the annual comprehensive cost by optimizing the planning schemes of distributed generation (DG), energy storage systems (ESS), and capacitor banks (CB); the lower level minimizes operating costs, voltage deviations, and carbon emissions by formulating operation strategies under typical scenarios, considering on-load tap changers (OLTC), controllable loads, capacitor banks, energy storage, and distributed generation. Then, the upper and lower levels of the model are coupled and unified into a single-level model. The normalized normal constraint (NNC) method is used to solve the single-level multi-objective model. Finally, simulation analyses are conducted on the IEEE 33-node distribution system to verify the model’s effectiveness.