Colorants have been a staple in the cosmetics industry for a considerable time, although certain varieties have been banned owing to health risks. Detecting and confirming these banned colorants simultaneously poses several challenges when employing LC–MS/MS. Molecular networking is a promising analytical technology that can be used to predict the structure of components and the correlation between them using structural and MS/MS spectral similarities. Molecular networking entails assessing the number of fragmented ions and the cosine score (the closer it is to one, the higher the similarity). In this study, we developed and verified a method for the simultaneous quantitative analysis of the 26 banned colorants in cosmetics using LC–MS/MS. Additionally, we propose a novel approach that combines LC–Q–TOF–MS and molecular networking technology to detect banned colorants in cosmetics. For successful molecular networking, a minimum of six fragment ions with cosine scores exceeding 0.5 is required. We developed a screening method for characterizing banned colorants using molecular networking based on LC–TOF–MS results for 26 banned colorants. Furthermore, we demonstrated that our established method can be used for screening by analyzing actual cosmetics (eyebrow tattoo, lipstick tattoo, and hair tint) spiked with three non-targeted banned colorants with similar structures (m/z 267.116, 315.149, and 345.157) in cosmetics. The combination of molecular networking techniques and LC–MS/MS proves highly advantageous for the swift characterization and screening of non-targeted colorants in cosmetics.