• The bioaccessibility of metals varied by source but remained stable across different particle sizes. • Cu, Mn, Cr, and V exhibited high bioaccessibility in fine particles due to combustion sources. • A bioaccessibility-based method was developed for source-specific risk assessment. • Explainable machine learning demonstrated that the aging process enhances the bioaccessibility of Mn and Cr. Metal bioaccessibility is critical for health effects. The influence of sources and aging processes on bioaccessibility of metals in size-resolved atmospheric particulate matter (PM) remains unclear. Size-resolved bioaccessibility of metals in source-emitted and atmospheric PM was measured in this study. The bioaccessibility of most metals varied greatly with sources, while it remained relatively stable across different sizes for each source. In atmospheric PM, Cu, Mn, Cr, and V showed high bioaccessibility at fine size but low at coarse size especially in floating dust season, indicating effects of combustion sources at fine size and dust sources at coarse size. Atmospheric As and Pb bioaccessibility showed slight size variability and were lower during heating season, linking with enhanced coal combustion with relatively low bioaccessibility. An advanced method was developed to quantify source-specific risk based on size-resolved bioaccessibility. Percentage contributions to cancer risk (CR) of PM 10 were the highest for industrial source (IS, 38 %), followed by aged aerosol (AA, 22 %), and coal combustion (CC, 18 %). Contribution of IS was high at sizes < 0.43 µm; and that of CC was high at sizes < 0.43 µm and 1.1–4.7 µm. Additionally, explainable machine learning revealed that atmospheric processes enhanced the Mn bioaccessibility, likely due to highly soluble MnSO 4 formed through acid-processing; and increased the Cr bioaccessibility, probably due to increased fractions of hexavalent Cr originating from oxidation processes.