<p>Acetone, as a major volatile organic compound (VOC), poses serious environmental and health risks, demanding efficient capture strategies. In this study, a robust framework that bridges classical adsorption isotherms with modern machine learning models is introduced to evaluate and predict acetone capture on carbonaceous adsorbents. A dataset of 1004 experimental points, collected from 28 porous carbons with diverse pore structures and surface chemistries, was analyzed using both classical and intelligent approaches. Three well-known isotherms—Langmuir, Freundlich, and Sips—were employed, and among them, the Sips isotherm provided the best description of adsorption behavior, achieving a sum of squared error (SSE) of 99.286. On the intelligent side, four models, namely, ANFIS, ANN, DT, and KNN, were applied, with the ANN model emerging as the most reliable, reducing the SSE to about 5.767, which is nearly an order of magnitude better than the best isotherm. The findings also highlight that the pressure, total pore volume, and BET surface area, with the respective Pearson correlation coefficient (PCC) values of 0.604, 0.564, and 0.547, are the most influential factors controlling the acetone uptake. While nitrogen content with the respective PCC value of -0.366 exhibits the most negative correlation. Collectively, the proposed intelligent models, with near-perfect predictive accuracy, demonstrate strong potential to replace classical isotherms in optimizing adsorbent performance. This research contributes to developing smarter carbonaceous adsorbents for effective VOC control, supporting public health and environmental sustainability.</p> Graphical Abstract <p></p>