Mechanical compatibility is a critical requirement for therapeutic patches applied to soft tissues and organs, where large deformation, nonlinear mechanical response, and auxetic surface behavior may occur. However, many current organ-patch design strategies rely on prescribed architectures, finite design libraries, or scalar mechanical descriptors, limiting their ability to match coupled tissue-like response profiles. In this study, we develop a diffusion-based multi-objective inverse design framework for organ-inspired auxetic metastructure patches that directly targets both stress–strain behavior and strain-dependent Poisson’s ratio. A finite element dataset containing 2669 parameterized unit-cell geometries from eight auxetic metastructure classes is first generated, with each sample labeled by both mechanical response curves. A convolutional neural network surrogate is trained to predict these responses from binary geometry images, achieving overall test-set R2 values of 0.9998 for stress and 0.9988 for Poisson’s ratio. An unconditional denoising diffusion probabilistic model is then trained to learn the feasible geometry distribution. During inverse design, the trained surrogate is embedded into the reverse diffusion process to provide response predictions and gradient information, thereby steering generated geometries toward prescribed target responses. To address the competing requirements of stress matching and Poisson’s ratio matching, an ε-constraint strategy is incorporated into the guided diffusion process to generate Pareto-compatible candidate sets representing different mechanical trade-offs. A finite element-based secondary screening step is further introduced to evaluate local strain-field restoration around a wound-like defect and select the final design according to an application-dependent repair criterion. A test-set verification case demonstrates that the framework can recover a known feasible geometry while also generating alternative designs with similar mechanical behavior, confirming the one-to-many nature of the inverse problem. Murine lung-inspired and human heart-inspired case studies further show that the proposed workflow can generate organ-inspired auxetic patch candidates and reveal different application-dependent preferences between stress matching and deformation matching. Overall, this work establishes a diffusion-based multi-objective inverse design framework that translates prescribed organ-inspired mechanical targets into final auxetic patch designs while directly considering coupled nonlinear response matching, stress–Poisson’s ratio trade-offs, and local deformation restoration.