<p>Contemporary research on microplastics primarily focuses on aquatic and atmospheric systems, thereby leaving the understanding of microplastic pollution in soils relatively underdeveloped. Conventional detection techniques often require labor-intensive preparation, costly instrumentation, and lack real-time capability. To address these limitations, we develop a laser ablation–proton-transfer-reaction mass spectrometry (LA-PTR-MS) method for rapid, sensitive detection of microplastics in soil. Polybutylene adipate terephthalate (PBAT) and polystyrene (PS) serve as representative biodegradable and conventional microplastics. Thermal degradation analysis identifies diagnostic volatile markers: PBAT generates propene (C₃H₆) and nonanal (C₉H₁₈O), while PS produces styrene (C₈H₈). These ions are accurately detected and show less interference from soil background. Multiple linear regression (MLR) models using these targeted markers achieve superior small-sample accuracy (n = 21) — PS: R² = 0.994, RMSE = 0.0729‰; PBAT: R² = 0.997, RMSE = 0.0527‰ — outperforming machine learning models built on full-spectrum signals because they reduce noise, focus on chemically specific features, and avoid overfitting. Field validation finds PBAT and PS concentrations of ~ 1 ‰ in plots with prior application, while other areas remain below detection limits. These results indicate that targeted-feature analysis not only enhances robustness and interpretability but also minimizes noise and overfitting in small-sample conditions. Nevertheless, full-spectrum data mining retains importance for leveraging microplastic fingerprint information. Together, these insights demonstrate, for the first time, that a laser-based PTR-MS platform integrated with chemo selective targeted regression can achieve reliable, preparation-efficient quantification, establishing a new paradigm for rapid soil microplastic monitoring.</p>