Three-dimensional quantitative structure–activity relationship (3D-QSAR) models were established by comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). Experimental toxicity data in Poecilia reticulata (pLC50) and physico-chemical properties for 12 polychlorinated phenols were used as dependent and as independent variables, respectively. Among the 12 polychlorinated phenols, nine were randomly selected and used as a training set to construct the 3D-QSAR models through the SYBYL-X software to predict the pLC50 values of the remaining 8 polychlorinated phenols congeners, and the other three polychlorinated phenols were used as a test set to evaluate the 3D-QSAR models (the training set and test set were arranged randomly, shuffled 60 times). Pentachlorophenol (PCP), which is the most toxic among the 20 polychlorinated phenols used in this experiment, was selected as an example for modification using contour maps produced using the established 3D-QSAR models. The aim was to decrease its toxicity and bioconcentration, increase its biodegradation, and maintain or better its effectiveness as a pesticide. The 3D-QSAR models were robust and had good predictive abilities with cross-validation correlation coefficients (q2) of 0.858–0.992 (>0.5), correlation coefficients (r2) of 0.966–1.000 (>0.9), and standard errors of prediction (SEP) of 0.004–0.159. CoMFA showed that the toxicity of the polychlorinated phenols arose mainly from electrostatic (42.7–66.7%) and steric (33.3–7.3%) contributions. By comparison, CoMSIA showed that the toxicity of polychlorinated phenols was dominated by electrostatic (57.5–76.9%) and hydrophobic (19.8–25.7%) contributions, with lesser contributions from the steric (0.7–1.0%) hydrogen bond donor (0.1–20.3%), and hydrogen bond acceptor (0–0.9%). 3D-QSAR electrostatic contour maps were used to modify PCP and design 11 new compounds with lower toxicity. The effectiveness of each of these molecules as a pesticide was verified using a 3D-QSAR model for polychlorinated phenol toxicity against Tetrahymena pyriformis. Four of these compounds, with −Br, −I, −OH and −NH2 groups in place of chlorine at the 3-position on PCP, were all at least as effective as PCP against T. Pyriformis. The first-order rate constants (Kb) of these four compounds were predicted using a 3D-QSAR model for polychlorinated phenol degradation, which showed they were more biodegradable than PCP. Furthermore, a 3D-QSAR model for polychlorinated phenols bioconcentration in fish (containing Poecilia reticulata, Oncorhynchus mykiss, Pimephales promelas and Oryzias latipes) showed that there was no significant difference between the bioconcentration factors of the four new compounds and that of PCP. The results obtained are hoped to provide a new route for lowering the POPs characteristics of those polychlorinated phenol homologues and derivatives in use.