The leaf area index (LAI) is an essential input parameter for quantitatively studying the energy and mass balance in soil-vegetation-atmosphere transfer systems. As an active remote sensing technology, light detection and ranging (LiDAR) provides a new method to describe forest canopy LAI. This paper reviewed the primary LAI retrieval methods using point cloud data (PCD) obtained by discrete airborne LiDAR scanner (DALS), its validation scheme, and its limitations. There are two types of LAI retrieval methods based on DALS PCD, i.e., the empirical regression and the gap fraction (GF) model. In the empirical model, tree height-related variables, LiDAR penetration indexes (LPIs), and canopy cover are the most widely used proxy variables. The height-related proxies are used most frequently; however, the LPIs proved the most efficient proxy. The GF model based on the Beer-Lambert law has been proven useful to estimate LAI; however, the suitability of LPIs is site-, tree species-, and LiDAR system-dependent. In the local validation in previous studies, poor scalability of both empirical and GF models in time, space, and across different DALS systems was observed, which means that field measurements are still needed to calibrate both types of models. The method to correct the impact from the clumping effect and woody material using DALS PCD and the saturation effect for both empirical and GF models still needs further exploration. Of most importance, further work is desired to emphasize assessing the transferability of published methods to new geographic contexts, different DALS sensors, and survey characteristics, based on figuring out the influence of each factor on the LAI retrieval process using DALS PCD. In addition, from a methodological perspective, taking advantage of DALS PCD in characterizing the 3D structure of the canopy, making full use of the ability of machine learning methods in the fusion of multisource data, developing a spatiotemporal scalable model of canopy structure parameters including LAI, and using multisource and heterogeneous data are promising areas of research.