In the Anthropocene, humans are the largest drivers of change in vegetation fire regimes. Humans influence fire regimes both directly, by starting, managing and extinguishing fires, and also indirectly by altering fuel composition and connectivity. However, whilst vegetation fire is a coupled socio-ecological process, representation of human influences on fire regimes in global-scale modelling remains limited. This places a fundamental constraint on our ability to understand how human and natural processes combine to create observed patterns of vegetation fire, and how such processes may interact under future scenarios of socioeconomic and environmental change. Here, we respond to this challenge by presenting a novel integration of two global and process-based models. The first is the Wildfire Human Agency Model (WHAM!), which draws on agent-based approaches to represent anthropogenic fire use and management. The second is JULES-INFERNO, a fire-enabled dynamic global vegetation model, which takes a physically-grounded approach to the representation of vegetation-fire dynamics. The WHAM-INFERNO combined model suggests that as much as half of all global burned area is generated by managed anthropogenic fires – typically small fires that are lit and then spread according to land user objectives. Furthermore, we demonstrate that including representation of managed anthropogenic fires in a coupled socio-ecological simulation can improve understanding of the drivers of unmanaged wildfires. Overall, findings presented here have substantial implications for understanding of present-day and future fire regimes, indicating that socio-economic change may be as important as climate change in determining the future trajectory of fire on Earth.