Agent-based modelling is a method for simulating urban systems from the bottom up. It allows for the simulation of human behaviours in certain urban environments, as well as the change of the urban environment caused by these behaviours. In this context, it has been widely used to investigate various problems pertaining to urban systems, such as urban land development, socio-spatial differentiation, and new town construction. A few literature reviews can be found, but they focus more on the principles, techniques, and deficiencies of ABM in studying urban systems, rather than a detailed and quantitative description of the research trajectories. Consequently, the application prospects of this method in urban systems are difficult to be further revealed. Against this backdrop, this study uses scientometric methods such as main path analysis and alluvial diagram analysis, as well as the data of literature citation network and keyword co-occurrence network to depict the research trajectories of agent-based modelling of urban systems. The result shows that land use is the primary focus of these studies. This research field consists of three stages: preliminary simulation of land use, the detailed simulation of the land market mechanism, and diversified applications. In the first stage, scholars began to use agent-based modelling to analyse urban land use, but the modelling approach was diverse and the rules for the agents’ behaviours were relatively simple. At the same time, they have yet not focused on the mechanism behind the land market. In the second stage, researchers focused more on the in-depth analysis and detailed simulation of the mechanism of the land markets. In the third stage, the studies have been becoming diversified again, focusing on the simulation of the land market, the reflection of agent-based modelling and the development of simulation systems for planners and the public. As for the research fields of agent-based modelling, recent literature from 2016 to 2019 focuses primarily on the following categories: Planning and industries, land use and urban growth, complexity and transportation, cellular automate and residential space, disaster and network, and others. These research fields are by no means fixed. Rather, they are highly flexible, intersecting with each other and having been undergoing continuous reconstruction. Among them, segregation and land use have received attention in early studies; urban growth, planning, transportation and disaster were studied in subsequent research; creative industries, accessibility, etc. are emerging research fields that received attention in recent years. In terms of the methods that have been combined with agent-based modelling, cellular automata, GIS and network analysis were combined with agent-based modelling in earlier studies. Besides, genetic algorithms, reinforcement learning models, and big data have also been used in agent-based simulation recently. In future research, studies on problems such as resilience city and the shrinking city can be integrated with agent-based modelling; some phenomenon specific to China, such as development zones and urban villages, can also be studied using this method. At the same time, future research can further focus on the combination of agent-based modelling with various artificial intelligence algorithms and traditional methods from pertinent subjects.