The Artificial Intelligence of Things (AIoT) connects everything with intelligence, while the increase in energy consumption generated by numerous electronic devices puts forward an impending demand on the power supply. Energy harvesting technology has emerged as a compelling innovation technology for the net zero emissions of the power supply for the AIoT. Although significant advances have been witnessed in energy harvesting, some issues such as poor electrical output, weak environmental adaptability, and low reliability are difficult to satisfactorily resolve. Mechanical intelligent energy harvesting can be defined as the system identifying the external excitation or its own state and reacting to it, rather than relying on electrical sensing elements or a central controller for certain adaptive or programmed functions. The adaptive and programmed functions exhibit great potential in solving the above‐mentioned issues that seriously restrict the development of the energy harvesting technology. Here, a generalized definition of mechanical intelligent energy harvesting is given critically and the design methodology is elaborated. The typical reported energy harvesting systems with the characteristics of mechanical intelligence are reviewed. The key research directions in mechanical intelligent energy harvesting are discussed. The mechanical intelligence is expected to revolutionize the development of the energy‐harvesting technology and pave the way for applications.