Illegal gold mining (IGM) is responsible for numerous environmental, social, and economic threats throughout the world. This is also the case in French Guiana (FG), where IGM causes impacts such as deforestation and the contamination of the food chain with methylmercury, among others. Since 2002, to address and mitigate these impacts, the French government has started taking actions and now treats IGM as a real matter of public policy and environmental governance. Despite the growing involvement of various public organizations and scientific communities, numerous challenges linked to the complexity of IGM in the middle of the Amazonian rainforest still result in a gap in applied scientific knowledge partially because of the difficulty to access the field. To improve long-term territorial planning in FG and optimize authority's efforts in terms of risk management, it is first necessary to better characterize and distinguish the different types of IGM practices, to determine their respective impacts on the territory, and to set action priorities. In FG, the existence of an observatory specifically devoted to IGM monitoring represents a unique opportunity to quantitively characterize and classify those activities. This paper presents an in-depth statistical analysis of a georeferenced database provided by this observatory, completed by individual interviews and sites visits. The result is a comprehensive classification of IGM practices in FG into five distinct types of illegal mining sites, based on six criteria (e.g., type of gold deposit, mining techniques), along with their detailed characteristics and inherent variabilities. The proposed classification is structured to distinctly differentiate the nature and the extent of the impacts of these five types of sites. Two of them are particularly detailed in this study, outlining their technical characteristics and their associated variability from site to site. The proposed classification and characterization of IGM in FG are essential for the development of realistic prospective territorial scenarios, and for their impact analysis. Although context-specific, the methodology proposed in this study has the potential to be applied to other countries facing similar IGM related issues, given the presence of equivalent databases or the willingness to create one.