FKA-Owl is a novel framework designed to enhance multimodal fake news detection by leveraging knowledge from Large Vision-Language Models (LVLMs) and incorporating forgery-specific knowledge. The framework addresses the challenge of detecting fake news, which often involves both text and image manipulations, by augmenting LVLMs with semantic correlation and artifact trace information. Two specialized modules are introduced: the cross-modal reasoning module and the visual-artifact localization module, which extract semantic correlations and artifact traces, respectively. These modules are then integrated into the LVLMs to improve their ability to reason about manipulations. Extensive experiments on public benchmarks demonstrate that FKA-Owl outperforms existing methods in cross-domain performance, showing superior results in single-domain, multiple-domain, and cross-dataset settings. The framework's effectiveness is attributed to its ability to effectively utilize world knowledge from LVLMs and the complementarity of the two specialized modules.FKA-Owl is a novel framework designed to enhance multimodal fake news detection by leveraging knowledge from Large Vision-Language Models (LVLMs) and incorporating forgery-specific knowledge. The framework addresses the challenge of detecting fake news, which often involves both text and image manipulations, by augmenting LVLMs with semantic correlation and artifact trace information. Two specialized modules are introduced: the cross-modal reasoning module and the visual-artifact localization module, which extract semantic correlations and artifact traces, respectively. These modules are then integrated into the LVLMs to improve their ability to reason about manipulations. Extensive experiments on public benchmarks demonstrate that FKA-Owl outperforms existing methods in cross-domain performance, showing superior results in single-domain, multiple-domain, and cross-dataset settings. The framework's effectiveness is attributed to its ability to effectively utilize world knowledge from LVLMs and the complementarity of the two specialized modules.