[Objective] Driven by the carbon peaking and carbon neutrality goals and energy transition， the large-scale grid connection of multiple entities such as high-share distributed generation， flexible loads， and new energy storage has promoted new distribution systems to exhibit characteristics including deep generation-network-load-storage coupling and dynamic evolution of morphological structures. Traditional experience-dependent planning models are no longer capable of addressing complex challenges such as the surge in system uncertainty and the coupling of multi-objective constraints. There is an urgent need to establish a new planning paradigm adapted to the development needs of the new distribution system by relying on artificial intelligence （AI） technologies. [Methods] This paper systematically sorts out the core processes and technical pain points of new distribution system planning， analyzes the key challenges in links such as planning knowledge graph construction， generation-load forecasting and scenario generation， power balance， capacity planning and network optimization， and generation-network-load-storage coordinated planning， and comprehensively elaborates on relevant research progress and technical application paradigms. This paper summarizes and analyzes the existing problems of AI-based new distribution system planning technologies， such as low efficiency in multi-source heterogeneous data fusion， insufficient cross-scenario generalization capability of models， lack of interpretability in planning decisions， difficulty in dynamic adaptation of multi-objective optimization weights， and inadequate consideration of the engineering feasibility of schemes. Corresponding key technical evolution directions are highlighted， including enhancing model adaptability through the integration of graph learning and physical constraints， strengthening cross-scenario and small-sample application capabilities via transfer learning and data augmentation， deepening data mining based on multi-modal fusion technology， improving decision interpretability by combining causal reasoning， and building a human-machine hybrid intelligence system to enhance planning practicality. [Results] Compared with traditional planning methods， AI technologies， relying on their powerful data processing and intelligent decision-making capabilities， provide efficient and accurate solutions for new distribution system planning， and have significant advantages in multi-scenario adaptation and multi-objective coordination. However， they still face a series of bottlenecks in technical implementation and engineering application. [Conclusions] In the future， it is necessary to continuously promote the deep integration of AI and new distribution system planning， tackle core technical challenges， and provide technical support and decision-making references for the high-quality development of distribution networks.