Digital twin (DT), aiming to characterise behaviors of physical entities by leveraging the virtual replica in real time, is an emerging technology and paradigm at the forefront of the Industry 4.0 revolution. The implementation of DT in predictive maintenance has facilitated its growth. As a major component of predictive maintenance, condition monitoring (CM) has great potential to combine with DT. To describe the state-of-the-art of DT-driven CM, this paper delivers a systematic review on the theoretical and practical development of DT in advancing CM. The evolution of concepts, main research areas, applied domains, and related key technologies are summarised. The driver of DT for CM is detailed in three aspects: data support, capability enhancement, and maintenance mode shift. The implementation process of DT-driven CM is introduced from the classification of DT modelling and the extension of monitoring algorithms. Finally, current challenges and opportunities for future research are discussed especially concerning the barriers and gaps in data management, high-fidelity modelling, behavior characterisation, framework standardisation, and uncertainty quantification.