Cyclooxygenase-2 (COX-2) is an inducible enzyme linked to inflammation and several diseases, including cancer and neurodegenerative disorders. Current synthetic COX-2 inhibitors carry serious risks of cardiovascular, gastrointestinal, hepatic, and renal issues. This study designed and validated four different predictive 2D- and 3D-QSAR models to assess the COX-2 inhibitory potential of chemicals, using selected molecular descriptors, in accordance with OECD guidelines. These models were used to screen a library of 54 phytocompounds and their analogues, and the results were further validated using ADMET screening, rodent toxicity prediction, molecular docking, and molecular dynamics simulations. The top-performing 2D models had R2=0.75, Q2=0.63, and pred_R2=0.51-0.53, while the 3D models had R2=0.64, Q2=0.55-0.69, and pred_R2=0.58-0.67, respectively. Of the four QSAR-prioritized compounds, apigenin-7-sulfate and genistein emerged as promising, toxicologically safe leads, with favorable binding affinity and stability within the COX-2 active site, compared to the well-known NSAID, celecoxib. Integrating multi-model QSAR, in silico pharmacology, and interaction analyses, this study provides insight into plant-based alternatives to COX-2 inhibitors and underscores the need for rigorous experimental/clinical validation in bioassays.