Two-dimensional transition metal dichalcogenides (2D-TMDs) with atomic thickness have attracted extensive attention due to their various physical properties, such as quantum spin Hall effect, superconductivity, charge density waves, ferroelectricity, and ferromagnetism. Owing to different interlayer stacking configurations and elemental coordination geometries, 2D-TMDs exhibit diverse crystalline phase structures with different physicochemical properties. Changing the crystalline phase structures of TMDs through phase engineering can be an effective strategy for modulating the electronic structures, quantum states, and functional characteristics. This review focuses on the manufacture of thermodynamically metastable-phase 2D-TMDs, providing a detailed discussion on the mechanisms of phase transition induced by physicochemical approaches and the latest advances in direct phase-selective synthesis of specific crystalline phase structures. The influences of phase engineering on electronic structures, superconductivity, magnetism, ferroelectricity, and other physical properties are systematically elucidated. The research advances in structure and property modulation of 2D-TMDs via phase engineering are summarized.At present, a variety of approaches including alkali metal intercalation, doping, defects, strain, electric field, and external stimuli (plasma, electron beam and laser irradiation) have been developed for controlled phase transition in 2D-TMDs. These physical and chemical approaches can induce local transitions of phase structure, which have the advantage of studying the process and mechanism of phase transition. However, there are still some problems such as the introduction of impurities and defects, insufficient phase stability, and challenges in large-scale fabrication. In contrast, the phase-selective synthesis of 2D-TMDs through methods such as temperature control, precursor design, interface engineering, seed crystal induction, and templated heteroepitaxial growth is more conducive to the characterization of intrinsic physical properties, large-scale fabrication, and electronic device applications. Despite the significant progress made in phase-selective synthesis, there are still several important challenges and development opportunities in this field. The general strategies and mechanisms of phase-selective synthesis still need to be further expanded and explored. In the future, it is expected that through theoretical simulations, machine learning-driven predictions and the integration of advanced in-situ characterization techniques, a universal and efficient phase engineering strategy will be developed, which can be extended to more 2D-TMD material systems.