<p>Nowadays, people spend a significant amount of time living and working indoors. For that, online detection and classification of different indoor air environments is of great importance, which is very challenging because of the lack of unified standards. In this study, a novel classification model based on laser-induced breakdown spectroscopy (LIBS) and machine learning was established to detect indoor air environments. To study different indoor air environments, four different coatings or paints (rubber coating, wood coating, furniture paint, and interior paint) were taken as samples to ascertain the gas composition. The analysis of their spectra shows there are various metal elements in these gas compositions, including Ti, Ca, and Na. For volatile organic compounds (VOCs) present in coatings or paints, the intensities of C, H, and O, which are VOCs’ main ingredients, are compared to determine if there is a certain difference. The results verify that LIBS could be used to detect different indoor air environments. Principal component analysis was used to distinguish the four indoor air environments, and the training data set was stored for further identification. Furthermore, a classification model was established based on the improved error back propagation artificial neural network (BP-ANN), achieving a recognition accuracy of 93.4%. After model training, the model's performance was tested using the spectra of different coatings or paints, and the recognition accuracy reached 98.2%. These results indicate that this method, combining LIBS and machine learning, has great potential for detecting the quality of the indoor air environments.</p>