The proliferation of multimedia content on social media has transformed how information is produced and consumed, enabling real-time coverage of global events but also accelerating the spread of misinformation, particularly during crises such as wars, natural disasters, and elections. The rise of synthetic media and the reuse of authentic content in misleading contexts further underscore the need for robust verification tools. In this paper, we present a comprehensive system developed for the ACM Multimedia 2025 Grand Challenge on Multimedia Verification. Our system evaluates both the authenticity and contextual accuracy of multimedia content in multilingual settings, producing expert-oriented verification reports alongside accessible summaries for the public. We propose a unified verification pipeline that integrates visual forensics, textual analysis, and multimodal reasoning, with a hybrid approach to detecting out-of-context (OOC) media via semantic similarity, temporal alignment, and geolocation cues. Extensive evaluations on the challenge benchmark demonstrate the system's effectiveness across diverse real-world scenarios. Our contributions advance the state of the art in multimedia verification while providing practical tools for journalists, fact-checkers, and researchers addressing information integrity in the digital age.

Misinformation is a widespread problem in the wake of flourishing social media. One of the most common forms of misinformation is cheapfakes - multimedia content that is manipulated with simple techniques such as miscaptioning to create misleading or false narratives. Due to their simplicity and appealing nature, cheapfakes pose a formidable threat to the reliability, transparency and integrity of the Internet. The ACM ICMR 2024 Grand Challenge on Detecting Cheapfakes tasks participants with detecting out-of-context cheapfakes - images with misleading or irrelevant captions. This paper introduces a lightweight approach that enhances cheapfake detection by combining Prompt Engineering with image captioning running on an interleaved image-text model. Testing on a public dataset gives an accuracy of 82.9%. Despite its modest accuracy, our method demonstrates the potential for applying better mixed-media learning models for context understanding and Visual Question Answering.