In shallow lake ecosystems, the exchange rate between water and sediment is high and can be easily disturbed by anthropogenic pressures. Additionally, catastrophic shifts occur when a threshold is reached; however, effective methods for detecting regime shifts can contribute to the timely control and restoration of eutrophication. Regime shifts of key variables, including chlorophyll, dissolved oxygen, and biomass of zooplankton and fishes, may be detected either through time series analysis, early warning signals provided by the abrupt changes of indicators, or identification of thresholds of explanatory variables, in which early warning signals can provide the most useful forecasting information. Thus far, the most widely used early warning indicators, such as autocorrelation and variance, are usually applied to lake ecosystems subject to the phenomenon of critical slowing down (CSD). However, under conditions of high stochasticity, strong external perturbation, and extreme events, these measures may underperform or subject to misinterpretation. Thus, in terms of driving mechanisms, the applicability and limitations of CSD indicators are herein discussed, as well as their prospects. The aim of this study is to synthesize what is currently known about the early warning detection of regime shifts in shallow lake ecosystems.